Multimedia Tools and Applications (2023) 82:43089–43144

https://doi.org/10.1007/s11042-023-15224-0

A Systematic survey on automated text generation tools

and techniques: application, evaluation, and challenges

Rupali Goyal1 · Parteek Kumar1 · V. P. Singh1

Received: 5 April 2021 / Revised: 2 March 2023 / Accepted: 30 March 2023 /

Published online: 19 April 2023
© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2023

Abstract
Automatic text generation is the generation of natural language text by machines. Enabling
machines to generate readable and coherent text is one of the most vital yet challenging
tasks. Traditionally, text generation has been implemented either by using production rules
of a predefined grammar or performing statistical analysis of existing human-written texts
to predict sequences of words. Recently a paradigm change has emerged in text generation,
induced by technological advancements, including deep learning methods and pre-trained
transformers. However, many open challenges in text generation need to be addressed,
including the generation of fluent, coherent, diverse, controllable, and consistent human-
like text. This survey aims to provide a comprehensive overview of current advancements
in automated text generation and introduce the topic to researchers by offering pointers
and synthesis to pertinent studies. This paper studied the relevant twelve years of articles
from 2011 onwards in the field of text generation and observed a total of 146 prime studies
relevant to the objective of this survey that has been thoroughly reviewed and discussed.
It covers core text generation applications, including text summarization, question–answer
generation, story generation, machine translation, dialogue response generation, paraphrase
generation, and image/video captioning. The most commonly used datasets for text gen-
eration and existing tools with their application domain have also been mentioned. Vari-
ous text decoding and optimization methods have been provided with their strengths and
weaknesses. For evaluating the effectiveness of the generated text, automatic evaluation
metrices have been discussed. Finally, the article discusses the main challenges and notable
future directions in the field of automated text generation for potential researchers.

Keywords Automatic Text Generation (ATG) · Text Generation Methods · Text Generation
Tools · Application-Specific Standard Datasets · Text Decoding Techniques · Automatic
Text Evaluation Methods

* Rupali Goyal
[email protected]
1
Computer Science and Engineering Department, Thapar Institute of Engineering and Technology,
Patiala, Punjab, India

13
Vol.:(0123456789)
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43090 Multimedia Tools and Applications (2023) 82:43089–43144

1 Introduction

Text Generation is a sub-discipline of Natural Language Processing used to fulfill specific

communicative requirements by automatically generating natural language texts that lever-
age computational linguistics and artificial intelligence abilities [168]. Text generation has
many real-world [51] applications depending on the input (data, text, or multimodal); how-
ever, the output is always a natural language text. Thus, based on the type of input, the text
generation has been categorized mainly into three categories: text-to-text generation (T2T),
data-to-text generation (D2T), and multimodality-to-text generation (M2T), as shown in
Fig. 1. The text-to-text generation tasks take existing text as input and automatically gen-
erate a new, coherent text as output. For T2T generation, the most common applications
include summarising the input document [13, 101], generating questions and answers
from a text document [4, 48, 193], translating a sentence from one language to another
[1, 11], creating or completing a story outline [66, 171, 211]. The data-to-text generation
tasks automatically generate text from numerical or structured data such as key-value lists
and tables. For D2T generation, the example applications include reports generation from
numerical data [148, 154], and generating text from the meaning representations to rep-
resent the meaning of natural language [113, 180]. The multimodality-to-text generation
tasks transfer the semantics in multimodal input, such as videos or images, into natural
language texts. For M2T generation, the example applications include generating captions
from images or videos [71, 126], video summarization, and visual storytelling.
The research on text generation has a long history. The earliest text generation systems
used template and rule based methods to capture linguistic knowledge of vocabulary, syn-
tax, and grammar. The next-generation models encode the dependency between vocabu-
lary and context in conditional probability. These methods also couple with template-based
methods for text generation. Then with the development of deep learning technologies,
neural-based models gradually occupy a dominant position. Deep Learning belongs to a
class of machine learning algorithms that identifies patterns in text and identifies features
that assist in solving several text generation tasks [97]. The capacity of deep neural net-
works to learn representations of varying degrees of complexity has aided in achieving
state-of-the-art performances across different text generation tasks, such as machine trans-
lation, text summarization, storytelling, and dialogue systems [62]. The availability and
accessibility of a vast number of corpora and massive computational resources are other
factors supporting deep learning growth. Most recently, the pre-trained text generation
models based on the Transformer architecture have the ability to better capture the lin-
guistic knowledge of vocabulary, syntax, and grammar. However, while these models gen-
erate fluent and grammatical text, they are prone to making factual errors that contradict
the input text. Generating fluent, informative, well-structured, and coherent text is pivotal

Fig. 1  Categories of common text generation tasks

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43091

for many text generation tasks. It takes significant effort by humans to generate text that
is consistent and model long-term dependencies. And it is an equally challenging to do
it automatically because of the discrete nature of textual data. Traditional template-based
methods generate reliable texts but lack diversity, fluency, and informativeness. The deep
learning models generate fluent and informative texts but are limited by the faithfulness
and controllability of the neural-based models. And the transformer based models generate
fluent, informative, and controllable text, but the inconsistency with the input information
persists. The motivation for conducting this survey and the primary contributions of the
paper are presented in the following subsection.

1.1 Motivation and our contribution

Automated text generation has been gaining attention with the advances in deep learning.
In the last ten years, the text generation field has evolved significantly. Numerous appealing
surveys [51, 59, 120, 123, 173, 216] have been introduced, summarizing the work done in
this field. However, there is no proper survey on ATG in terms of prominent benchmark
datasets, real-world tools, decoding methods, evaluation metrices, and challenges of auto-
mated text generation applications. This motivates us to perform a Systematic Literature
Review for automatic text generation using deep learning techniques. Another motivation
is the gaining interest in this research field of text generation. The analysis of the twelve-
year articles published in this field has consistently increased, indicating that automated
text generation is gaining interest each year. Keeping this in mind, this paper studies all the
relevant articles from 2011 onwards to find methods for automated generation of text in dif-
ferent application domains, different existing tools and datasets used with their application
domain, and evaluation metrices for evaluating the effectiveness of the generated text. The
purpose of this survey is to provide a comprehensive overview of current advancements in
automated text generation and introduce the topic to researchers by providing pointers and
synthesis to pertinent studies.
The main contribution of this paper comprises the following key points:

I. This survey provides an up-to-date synthesis of automated text generation along with
its core applications, including text summarization, question–answer generation,
dialogue generation, machine translation, story generation, paraphrasing, and image
captioning, and the key techniques behind them.
II. A comprehensive outline of techniques and methods employed to generate text auto-
matically, including traditional statistical methods, deep learning, and pre-trained
transformer based models, has been discussed.
III. This paper enlists standard datasets required to train, test, and validate the text gen-
eration models for the automated generation of fluent and coherent texts.
IV. The text decoding strategies and optimization techniques significantly impact the
quality of the generated text. This paper discussed these decoding techniques and
optimization methods with their strengths and limitations.
V. In this article, real-time task-specific tools for automated text generation have also
been provided with their features and URLs.
VI. For the effectiveness of the generated text, various metrices/approaches have also
been summarized to evaluate text generation models automatically that depict dif-
ferent text attributes such as fluency, grammaticality, coherence, readability, and
diversity.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43092 Multimedia Tools and Applications (2023) 82:43089–43144

VII. This paper also identifies various challenges of text generation applications, including
the generation of human-like text that is fluent, diverse, controllable, and consistent.
This survey also outlines potential future research directions in the area of automated
text generation.

1.2 Comparison with other surveys

There have been related attempts and literature surveys on automated text generation. This
subsection overviews such attempts and highlights the contrast between existing and cur-
rent surveys with their strengths and limitations. For example, Gatt et al. [51] surveyed
natural language generation emphasizing image-to-text tasks. Liu et al. [120] reviewed
deep learning architectures and limited text generation applications. Xie [216] describes
techniques for training and dealing with natural language generation models using neural
networks. Santhanam et al. Lu et al. [123] surveyed only neural text generation models.
[173] review language generation with a focus on dialogue systems. Garbacea et al. [59]
present an overview of natural language generation methods, tasks, and assessments. Yu
et al. [232] reviewed knowledge-enhanced text generation. However, there are many exist-
ing surveys on text generation but are limited in terms of standard datasets, existing real-
time tools, optimization methods, evaluation metrices, and challenges of automated text
generation applications. This motivates us to perform a Systematic Literature Review on
automatic text generation. This survey captures the comprehensive study and up-to-date
review of current advancements in the field of text generation and also studies various
methods for automated generation of text in different application domains, different exist-
ing tools, and datasets used with their application domain, text decoding, and optimization
techniques, and evaluation metrices for evaluating the effectiveness of the generated text.
Table 1 summarizes these aspects in comparison to the surveys mentioned above reports
and contributions in literature.
The rest of this paper is organized as follows. A detailed review strategy and various
research questions with significance are provided in Sect. 2. This section also mentions the
search criteria and the research parameters for writing this survey paper. The extraction
of studies and discussion is presented in Sects. 3 to 10. In Sect. 3, the core applications of
text generation are reviewed. Section 4 mentions the methods and techniques employed for

Table 1  Comparative analysis of Authors [Ref.] 1 2 3 4 5 6 7

the proposed survey with existing
surveys
Gatt et al. [51] ✓ ✓ ⋆ ⋆ ✗ ⋆ ✗
Liu et al. [120] ⋆ ✓ ✗ ✗ ✗ ✗ ✗
Xie [216] ✗ ✓ ✗ ✗ ✗ ✗ ✗
Lu et al. [123] ✗ ✓ ✗ ✗ ✗ ✗ ✗
Santhanam et al. [173] ✓ ✓ ✗ ⋆ ⋆ ✗ ✗
Garbacea et al. [59] ✓ ✓ ✗ ✗ ✗ ✓ ✗
Yu et al. [233] ✓ ✓ ✗ ✗ ✗ ✗ ✗
Our Survey ✓ ✓ ✓ ✓ ✓ ✓ ✓

where 1: Text Generation Applications, 2: Text Generation

Approaches, 3: Real-Time Tools, 4: Standard Datasets, 5: Text Decod-
ing and Optimization Methods, 6: Automatic Evaluation Metrices, 7:
Text Generation Application Challenges; and✓—Detailed Study, ⋆—
Limited Consideration, ✗—No Discussion.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43093

generating text. Section 5 enlists application-specific standard datasets required to train,

test, and validate the models. Text decoding and optimization techniques significantly
impact the generated text, which are mentioned in Sect. 6. There are many real-time task-
specific tools for text generation, which are provided in Sect. 7, along with their access
URLs. Section 8 reviewed the approaches to evaluate the effectiveness of the generated
text. Various open challenges to automate the text generation task is mentioned in Sect. 9.
Finally, Sect. 10 concludes this paper and outlines potential directions for future research.

2 Research methodology

The research methodology is a process of systematically researching. It includes an empiri-

cal analysis of all concepts relevant to the field of research. Generally, it includes the con-
cepts of phases, models, and quantitative as well as qualitative techniques. This paper
follows the review process suggested by Kitchenham and Charters [92], which includes
planning, conducting, and reporting the review, as shown in Fig. 2.

2.1 Planning review

The planning process included identifying the need for a Systematic Literature Review
(SLR) and concluding with the formulation and validation of the review procedure. A
systematic review is needed to identify, compare, and classify the existing text generation
work. The studies published on text generation in the last twelve years are observed, but
none are robust. This paper comprehensively analyzes emerging models, methods, tools,
and deep-learning-based text generation application techniques to identify and compare
them systematically.
The research questions (RQs) are prepared to facilitate the review process to be more
focused, clear, and consistent. Eight research questions (RQ 1 to RQ 8) have been framed,
which help to perform SLR. The research questions and their significance in this literature
review are mentioned in Table 2.

Fig. 2  Overview of research methodology

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43094 Multimedia Tools and Applications (2023) 82:43089–43144

Table 2  Research Questions and their Significance

Research questions Significance

RQ 1: How the automated text generation study To identify the rate of adaptation of automated
evolves with advancements in deep learning? text generation studies and the yearly analysis of
research in text generation with advancements in
the technology models
RQ 2: What are the main text generation applica- To classify text generation tasks and to describe
tions that have arisen with advancements in the various text generation applications based on input
field? and usage
RQ 3: What are the various approaches and associ- To identify and analyze various methods and tech-
ated architectures in the field of text generation? niques for automatic text generation. This helps to
identify strengths and limitations associated with
the approaches
RQ 4: What are the available datasets adopted in To identify the available standard datasets according
which the stated applications are organized? to application to test, train and validate the model
RQ 5: What are various text decoding and optimiza- To identify the task specific text decoding strategies
tion techniques are used to generate fluent text and optimizers for automated text generation
automatically?
RQ 6: What real-time tools are available for auto- To have a perception of the usage of text genera-
matic text generation tasks? tion applications in the real world and to analyze
the strengths and weaknesses of the existing text
generation tools
RQ 7: Which metrics or indicators are used to evalu- To evaluate the effectiveness of the generated text,
ate the generated text? identification of the evaluation metrices used for
text generation applications
RQ 8: What challenges are faced in automated text To identify the issues related to the applications of
generation tasks? automatic text generation tasks

2.2 Conducting review

This phase involves selecting studies, extracting required resources, and synthesizing
knowledge. This SLR includes research papers from different publications and the var-
ious online electronic databases selected, such as IEEE Xplore, ACM digital library,
Science Direct-Elsevier, Springer link, Web of Science, and Wiley online library. The
search string includes keywords: “Text Generation” OR “Natural Language Generation”
OR “Text Generation using Deep Learning” OR “Neural Text Generation” OR “Neural
Language Generation” AND “Applications” OR “Text Generation Applications.” The
sources contain documents of several types, such as book chapters, research articles,
reviews, and proceeding papers, published in the last twelve years, i.e., from 2011 to
2022. It discusses the research papers from journals, magazines, conferences, work-
shops, and symposiums. The studies were explored and based on inclusion–exclusion
criteria, and a total of 146 research papers were obtained, as shown in Fig. 3.
These 146 research papers from the ‘2011–2022’ time frame are thoroughly reviewed
and discussed in this survey paper. The number of extracted research papers based on
their year of publication is shown in Fig. 4. It can be observed that before 2011 there
was limited work in the research area of text generation using deep neural networks.
And there has been a gradual increase in the number of research papers from 2011
onwards, showing growth in the field of automated text generation with the develop-
ments in deep neural models.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43095

Fig. 3  Inclusion/Exclusion technique used in the systematic review

35
30
25
# Papers

20
15
10
5
0
2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022
Year

Fig. 4  Yearly analysis of the papers in the text generation research area with development in technology

Discussion on research question 1

This yearly analysis of papers helps answer RQ 1- “How the automated text generation
study evolves with advancements in deep learning.” It has been observed that limited work
has been done on text generation before 2011, and there is a gradual increase in the number
of research papers from the year 2011 onwards. This growth results from advancements in
text generation methods, from traditional rule-based methods to deep neural networks and
pre-trained transformer models. The traditional template or rule based methods were used
for text generation usually before 2013, but these rules/templates are difficult to design and
are very time consuming. These shortcomings of traditional approaches were overcome
in the years with the developments in deep learning methods. The research in the field
of text generation increases gradually thereafter. The availability of powerful deep neural
models and computationally intensive architecture results in the incredible adoption of a
variety of text generation applications, including text summarization, machine translation,

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43096 Multimedia Tools and Applications (2023) 82:43089–43144

Fig. 5  Reviewing parameters for each text generation application

creative applications such as story generation, and dialogue generation. Now, with pre-
trained transformers models, the automated text generation work has immensely acceler-
ated. Many sectors have started using automated text to improve user experience as the
recent advancements in technology is capable of generating human-like texts. The content
generated by the automated tools is fast and cheap. The analysis of the published articles in
this field indicates consistent growth and adaptation to the research area of automated text
generation from 2011. Keeping this in mind, this paper studies all the relevant articles from
2011 onwards to find methods for automated generation of text in different application
domains, different existing tools and datasets used to achieve the task, text decoding and
optimization techniques, and evaluation metrices for the effectiveness of the generated text.

2.3 Reporting review

For reporting the review, this phase provides the research parameters. The research param-
eters that have been followed for SLR include the core text generation applications. For
each text generation application, the deep learning technique used, standard datasets, exist-
ing real-time tools for that application, optimization methods, evaluation metrics, and chal-
lenges for each application are presented. The different parameters used in this paper are
shown in Fig. 5. The extraction of studies and discussions is presented in this survey.
These research parameters are discussed in-depth in the following sections, and the
results or analysis of these parameters are also reviewed and mentioned in Sects. 3 to 10.

3 Applications of automated text generation

The field of artificial intelligence has developed techniques that generate text automatically
in seconds. Automatic text generation is one such application that is the need for the hour.
Many applications of text generation are crucial and very significant for smart systems and
enable better communication between humans and machines, for example, machine trans-
lation, summarization, and simplification of long or complex texts, grammar, and spelling
correction, generating peer reviews for scientific papers, questionnaire generation, auto-
matic documentation systems for large software, question–answer generation, business let-
ter writing, chatbots and much more. The core text generation applications are shown in
Fig. 6, and the details about these applications are discussed below.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43097

Fig. 6  Text Generation Applications

3.1 Text summarization

With each day, enormous amounts of data from diversified sources have evolved. This mas-
sive volume of data incorporates crucial facts, information, and knowledge that needs to
be effectively summarized to be helpful. Thus, automatic text summarization came into
the picture to tackle the problem of information overloading [142]. A text summarization
method generates an abbreviated version of a document by filtering the significant informa-
tion from the original document [57]. A strong summary consists of all aspects, such as
coverage, non-redundancy, cohesion, relevancy, and readability, in addition to relevant key
points [145]. There are two prominent types of summarization techniques. First, extractive
summarization techniques form summaries by copying parts of the input sentences [134],
and second, the abstractive summarization technique [5, 132] generates a summary by
including words and phrases not present in the source [135]. Nallapati et al. [135] propose
recurrent neural network-based encoder-decoder models for abstractive text summarization.
In follow-up work [134], extractive summarization techniques using recurrent neural net-
works are presented. Rush et al. [172] propose an attention-based network for the abstrac-
tive summarization of sentences, and Cheng et al. [28] proposed an attentional encoder-
decoder for extractive single-document summarization. See et al. [177] used a pointer
generator network for abstractive summarization. Paulus et al. [147] use the reinforcement
learning model for abstractive summarization, while others use reinforcement learning for
extractive single-document summarization [136, 212]. Mehta et al. [129] use Long Short
Term Memory (LSTM) and attention model to summarize scientific papers. Liu et al. [118]
focus on multi-document summarization by generating fluent, coherent multi-sentence
Wikipedia articles using extractive summarization. Modified BERT transformer [40] for
extractive summarization is capable of extracting automatically the features in the internal
layers [116] Multi-document summarization using abstractive methods has also been used
[15, 239]. Xu et al. [221] propose a multi-task framework with a hybrid of the extractive
and abstractive models. Transformer architecture also performs great in many NLG tasks
[195]. Tan et al. [20] used a pretrained model, GPT-2 for the summarization task with the
idea that the model will start generating a summary based on the delimiter. More recent
works leverage pre-trained transformer based networks, such as GPT [162], BART [102],
T5 [163], and PEGASUS [240], for summary generation [63, 119, 213].

3.2 Question answer generation

Automatic question generation (QG) aims to generate questions from some form of input,
such as raw text or a database, whereas Question Answering (QA) is the task of automati-
cally providing precise responses to questions in the natural language given corresponding
document. In the last years, the widespread use of QA-based personal assistants has been
observed, including Microsoft’s Cortana, Apple’s Siri, Samsung’s Bixby, Amazon’s Alexa,

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43098 Multimedia Tools and Applications (2023) 82:43089–43144

and Google Assistant, which have answered a wide variety of questions. QG systems pro-
posed by [46, 185] automatically generate answer-unaware questions from within the given
document, whereas [88, 180, 186] generate answer-agnostic questions. Du et al. [47] initi-
ated a neural question generation model using an attention sequence-to-sequence model
[11]; subsequently, [48, 69, 245] also adopted an attention mechanism. Zhao et al. [241]
proposed a gated self-attention encoder. Most neural QG models [69, 95, 204, 245] employ
the copying mechanism for question generation. Weston et al. [206] proposed the use of a
Memory networks model in the system to answer the questions effectively. The Dynamic
Memory Networks model [94] overcomes the shortcomings of the memory networks by
combining the paradigms of memory networks and attention mechanisms. This work was
later extended by Xiong et al. [219] for visual question answering. Other works, includ-
ing visual question answering [2, 9, 58, 122] have generated natural and engaging ques-
tions for an image. [233] have adopted policy gradient methods to diversify the generated
question. [40, 99, 225] uses pre-trained models for the question-answering task, and [100]
uses transformer-based models to generate answer aware questions. [203, 204] propose a
neural model for question generation and answering that jointly asks and answers ques-
tions given a document. Most of the earlier work focuses on using a single QA dataset,
such as SQuAD [165]. While working on the generation of multi-hop [30], open ended and
controllable [23], or cause-effect [183] questions have gained attention, each direction is
studied in isolation as it usually requires a separate question–answer dataset. More recent
works leverage pre-trained transformer based networks, such as BART [102], T5 [163],
and PEGASUS [240], for question generation, which have been successful in many appli-
cations [6, 98, 107, 164, 194].

3.3 Dialogue response generation

Dialogue systems or conversational agents are computer programs capable of replying with
natural, coherent, meaningful, and engaging responses. A good dialogue model generates
dialogues with high human similarity [104]. [131, 181] work on building end-to-end dialog
generation systems using neural networks, whereas [178, 191] use hierarchical encoder-
decoder to generate responses. [218, 228] use the attention model, while [103, 110] use rein-
forcement learning, and Li et al. [111] use generative adversarial networks for dialog gen-
eration. Niu et al. [141] also use a reinforcement learning model focusing on polite dialogue
responses. The use of pre-trained models for conversational agents is also observed. [10,
242] use embeddings; [41, 209] use transformers for response generation [98, 107]. These
conversational models have enabled robots to interact with humans in natural languages; for
example, Window’s Cortana, Google’s assistant, Apple’s Siri, and Amazon’s Alexa are the
software and devices that follow Dialogue systems. [76, 153] proposes dialogue generation
with recognition of emotions, and [56, 167] also generates empathetic dialogues. [181, 243,
244] generates single-turn dialogue responses while [159, 167, 235] generates multi-turn dia-
logues. [184] has used text style transfer and GPT for the creation of a dialog generation sys-
tem over gender-specific, emotion-specific, and sentiment-specific dialogue datasets.

3.4 Neural machine translation

The data accessible to everyone is a challenge because language becomes the barrier, and
machine translation came into the picture to overcome this. Machine translation is the task

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43099

of automatic translation of written text from one natural language into another. Neural
machine translation (NMT) uses neural nets to transform the source sentence into the target
sentence [29, 82, 188]. [11] introduced attention mechanism in NMT models, which was
later extended [124]. Luong et al. [124] have used a unidirectional recurrent neural network
model, while [214] used bidirectional recurrent neural networks (BRNN). [3] and [79]
does the translation for low-resource language pairs. [29, 31] uses gated recurrent units
and achieves better performance on NMT. Tu et al. [192] use a copy and coverage mecha-
nism; Wang et al. [200] use a pre-computed word embedding layer, GlOVe (Global Vectors
for Word Representation). Park et al. [144] proposed a mobile device-based sign language
translation system. [1] uses an attention-based multi-layer neural network. Transformer
architecture [195] also performs well in NMT. [12, 200] proposed deep Transformer mod-
els for translation. Transformers like BART [102], BERT [35, 40] and GPT [20] have also
been used for the NMT task. More recent works leverage pre-trained transformer based
networks for machine translation [54, 64]. Recent works on non-autoregressive neural
machine translation [65, 72, 161] improve model efficiency by decoding in parallel as com-
pared to sequential decoding in traditional autoregressive machine translation methods
[215].

3.5 Story generation

Automated story generation is the task of automatically identifying a series of actions,

events, or words that have been told as a story. Li [106] attempts to automatically generate
a story about any domain without prior knowledge. To encode the context, recurrent net-
works, and convolutional networks successfully model sentences [38, 81]. A fusion mech-
anism [182] is introduced to support sequence-to-sequence models build dependencies
between their input and output. Pawade et al. [149] have implemented a recurrent neural
network-based story system to generate a new story based on a series of inputted stories.
Vaswani et al. [195] use multi-head attention. [108, 170] use LSTM networks to learn the
text hierarchically. Jain et al. [78] chain a series of variable length independent descriptions
together into a well-formed comprehensive story. Clark et al. [34] model entities in story
generation. Martin et al. [128] present an event-based end-to-end story generation pipeline.
Similarly, [68] generates summaries of movies as sequences of events using a recurrent
neural network (RNN) and sample event representations. [53, 227] propose a hierarchical
story generation framework that first plans a storyline and then generates a story based
on the storyline. [151, 208] propose a framework that enables controllable story genera-
tion. [7, 189] uses policy gradient deep reinforcement learners to perform an event-to-event
task. [26] uses the BERT language model for story plot generation.

3.6 Paraphrase generation

Texts that convey a similar meaning but different expressions are referred to as para-
phrases. Paraphrase generation refers to an activity in which, given a sentence, the system
creates paraphrases of it. Bowman et al. [18] use a variational autoencoder (VAE) to model
holistic properties of sentences such as style, topic, and other features. Gupta et al. [67] use
VAE-LSTM to generate more diverse paraphrases. Prakash et al. [155] employ a stacked
residual LSTM network in the Sequence-to-sequence model. [105, 166] propose deep rein-
forcement learning (RL) to guide Sequence-to-sequence training. Cao et al. [22] utilize a
novel sequence-to-sequence model to join copying, and restricted generation [237] tackle

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43100 Multimedia Tools and Applications (2023) 82:43089–43144

a comparable task with the Sequence-to-sequence model coupled with deep reinforcement
learning. See et al. [177] use a pointer-generator while [125] utilizes an attention layer.
Iyyer et al. [77] utilize syntactic information for controllable paraphrase generation. Yang
et al. [224] propose an end-to-end conditional generative architecture for generating para-
phrases. Qian et al. [160] propose an approach that generates a diverse variety of differ-
ent paraphrases. [21, 43] tackle the problem of QA-specific paraphrasing while [223] help
diversifies the response of chatbot systems. [117] first uses abstract rules and then leverage
neural networks to generate paraphrases by refining the transformed sentences.

3.7 Image/Video caption generation

The generation of semantically and syntactically correct description sentences of an image

is called image captioning. The recognition of vital objects, their properties, and their
relationships in an image is required for image captioning. Kiros [90] propose the initial
work for extracting image features with the use of a convolutional neural network (CNN)
in generating image captions. Then, with the use of LSTM [90] extended their work [91].
Mao et al. [127] proposed a multimodal recurrent neural network (m-RNN) and [229] used
hierarchical recurrent neural networks for generating image descriptions. [201] proposed
a deep Bi-LSTM based method for image captions. [80, 220] proposed an attention-based
image captioning method. [169, 238] introduced a reinforcement learning-based image
captioning method. [37, 179] proposed an image captioning method based on Generative
Adversarial Networks (GAN). Vinyals et al. [199] proposed a neural image caption genera-
tor method. Donahue et al. [42] propose long-term recurrent convolutional networks that
have been processing variable-length inputs. [150, 230] propose an attention-based image
captioning model. Some method uses a CNN for image representations and an LSTM for
generating image captions. Yao et al. [226] proposed a copying mechanism to generate a
description for novel objects. [55, 85] use pre-computed word embedding layers and thus
generate better image captions. [202] has proposed a framework that unifies a diverse set
of cross-modal and unimodal tasks, including image captioning, and language modelling.

Discussion on research question 2

Text generation research consists of various tasks, topics, or trends. This section helped
to answer RQ 2- “What are the main core text generation applications that have arisen with
advancements in the field.” It has been observed that there are many real-world applica-
tions depending on the input (data, text, or multimodal); however, the output is always a
natural language text. Thus, based on the type of input, the text generation has been cat-
egorized mainly into three categories: text-to-text generation (T2T), data-to-text generation
(D2T), and multimodality-to-text generation (M2T), as discussed earlier. For T2T genera-
tion, the most common applications include summarising the input document, generating
questions and answers from a text document, translating a sentence from one language to
another, and creating or completing a story outline. For D2T generation, the example appli-
cations include reports generation from numerical data and generating text from the mean-
ing representations to represent the meaning of natural language. For M2T generation, the
example applications include generating captions from images or videos, video summari-
zation, and visual storytelling. The summarized description of the above-mentioned text
generation applications is mentioned in Table 3.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Table 3  Summarized description of text generation applications
Application [Ref.] Description Input Output

Machine Translation [11] Translate a text document from a source A text document in language X (e.g., Punjabi) Document in language Y (e.g., English)
language into a target language
Text Summarization [135] Summarize the text document into a concise A long text document like research work A summary
and shorter text
Headline Generation [121] Summarize an article into a brief headline A long document like a news article A headline
Question Answering [50] Given a question and a text document, gener- A textual question A short text as an answer from the document
ate the answer to the question
Question Generation [52] Generate questions from a textual document, A text document, article, or image A wide range of questions related to the input-
article, or an image ted document
Multimedia Tools and Applications (2023) 82:43089–43144

Dialogue Generation [10] Generate dialogue among agents, e.g., A dialogue from the audibles or the first agent A dialogue from the audibles or the second
between a human and a robot agent
Story Telling [170] Given an image or story outline, generate a A story outline or an image A story in text form
textual story
Smart Reply/email answer Given an original message, suggest the most An incoming email/message Most probable responses
suggestion [84] likely responses
Paraphrase Generation [155] Given a sentence, generate multiple para- An original sentence Multiple paraphrases for the given sentence
phrases
Image Captioning [201] Generate a caption that explains the image’s An image A caption describing that image
content
Video Description [112] Generate a video description that explains the A video clip A sentence describing that video
content of the video

13
43101

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

43102 Multimedia Tools and Applications (2023) 82:43089–43144

The availability of powerful deep neural models, computationally intensive architecture,

and pre-trained transformer models results in the incredible adoption of a variety of text genera-
tion applications. The applications use different methods including recurrent neural networks, long
short-term memory networks, gated recurrent units for learning language representations, and later
sequence-to-sequence learning, which opens a new chapter characterized by the wide application
of the encoder-decoder architecture. However, these sequence-to-sequence models cannot capture
long term dependencies, motivated the development of pointer networks and attention networks.
Then, the transformer architecture incorporates an encoder and a decoder with self-attention mech-
anism, which is now widely used by text generation tasks. Applying these models to different text
generation tasks can result in different levels of performance due to differences in task-specific
requirements, training data availability, model architecture, hyperparameters, and evaluation met-
rics. Even if the same or similar models are used for different tasks, the architecture of the model
may need to be modified or fine-tuned based on the requirements of the specific task.
The availability and quality of training data can significantly impact the performance of a text
generation model. Models that are trained on large, diverse, and high-quality datasets specific to a
given task or domain tend to perform better than those trained on more general datasets or limited
data. Different text generation tasks may require different data preprocessing steps, such as tokeni-
zation, normalization, stemming, and stop-word removal. The choice of the model architecture and
hyperparameters can also impact the performance of a text generation model. For example, trans-
former-based models such as GPT tend to perform well on a variety of text generation tasks due to
their ability to capture long-term dependencies, but different hyperparameters, such as the number
of layers or attention heads, can affect the model’s performance. Different text generation tasks have
different requirements and constraints that affect the effectiveness of the model. Models that are
optimized for a specific task may perform better than those that are more general-purpose. Thus, it
is important to carefully consider these factors when selecting a model for a particular task.
In this section, advancements in text generation applications have been seen with the rise of deep
neural network approaches. The text generation approaches are discussed in the next section.

4 Text generation approaches

Text Generation is an emerging area of research. Recently, deep learning approaches have
made remarkable success in various text generation tasks [138], including text summari-
zation, machine translation, question answering, story generation, short-dialog generation,
and paraphrasing. This section presents the traditional approaches for text generation, deep
learning techniques, and pre-trained transformer-based approaches to text generation.

4.1 Traditional approaches

Traditionally, text generation was done either by using templates or production rules of a
predefined grammar or performing statistical analysis of existing human-written texts to
predict sequences of words [17, 60, 139, 222]. The template-based text generation sys-
tems adopted rules and templates to design different modules for text generation that reflect
the linguistic knowledge of vocabulary, syntax, and grammar. This approach decomposes
the text-generation task into several interacting subtasks depending on the task-specific text
generation application. The template-based approaches usually consist of several compo-
nents, including content planning (deciding the input data, selecting and structuring content),

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43103

sentence planning (choosing words, syntactic structures, choosing appropriate referring

expressions to describe input entities), and text realization (converting specifications to a real
text), each performing a specific function [137]. The statistical-based text generation systems
encode the dependency between vocabulary and context in conditional probability [93]. The
most popular statistical text generation model is the n-gram language model, which is usually
coupled with the template-based approach for re-ordering and selecting fluent generated texts.
With the traditional approaches, it is very time-consuming to automatically generate text like
those generated by humans. Deep learning techniques have overcome these shortcomings of
traditional approaches. With the development of deep learning approaches, the neural-based
text generation models have gradually occupied a dominant position that better models the
statistical relationship between vocabulary and context, thus significantly improving the per-
formance of text generation, as discussed in the subsequent section.

4.2 Deep learning techniques

Deep learning architectures and algorithms have recently achieved state-of-the-art

results in question–answer generation, machine translation, text summarization, dia-
logue response generation, and other text generation tasks. Deep learning supports
automated multi-level attribute representation learning. The deep neural networks pro-
vide a uniform end-to-end framework for text generation. First, a neural network cre-
ates a representation of the user input. Then, this representation is used as input to a
decoder which generates the system response. Representation learning often happens
in a continuous space, such that different modalities of text (words, sentences, and
even paragraphs) are represented by dense vectors.
A variety of architectures based on deep neural networks have been developed for the
different application tasks of text generation. This section introduces deep learning tech-
niques that are commonly used in text generation application tasks.

4.2.1 The encoder‑decoder framework

Much of the work on neural text generation adopts the encoder-decoder approach that was
first advocated and shown to be successful for machine translation [32, 188]. First, the input
is encoded into a continuous representation using an encoder. Then the text is produced using
the decoder. Figure 7 illustrates this encoder-decoder framework for text generation. This net-
work is often referred to as a sequence-to-sequence model as it takes as input a sequence, one
element at a time, and then outputs a sequence, one element at a time.
This encoder and decoder are neural networks. The encoder depicts the input sequence
as a hidden state vector and then transfers it to the decoder. The decoder then produces the
output sequence.

4.2.2 Convolutional neural networks

Convolutional neural networks (CNNs) are specialized for processing data with a known
grid-like topology. These networks have succeeded in computer vision tasks, which have
been represented as 2-dimensional grids of image pixels [201, 217]. In recent years, CNNs
have also been applied to natural language. In particular, they have been used to learn word

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43104 Multimedia Tools and Applications (2023) 82:43089–43144

Fig. 7  Encoder-Decoder frame-

work

Fig. 8  Convolutional Neural Network

representations for language modelling effectively [177] and summarization [28, 136,
138]. CNNs employ a specialized kind of linear operation called convolution (filter which
extracts a specific pattern), followed by a pooling operation (subsamples the input on each
filter to a fixed dimension of output), to build a representation that is aware of spatial inter-
actions among input data points as shown in Fig. 8.
There are many variants of CNN that have different application areas, as mentioned in
Table 4.

4.2.3 Recurrent neural networks

Recurrent Neural Networks are based on the concept of processing sequential data. They
are termed “recurrent” since they perform the same computation over each token in the
sequence, and each step depends on the results of previous computations, as shown in
Fig. 9.
Most work on neural text generation has used RNNs due to their ability to capture the
sequential nature of the text naturally and to process inputs and outputs of arbitrary length.
There are various variants of RNN, including Bi-directional RNN, Parallel-RNN, Quasi
RNN, RNN with external memory, and Convolutional RNN. Their features and application
areas are provided in Table 5.
However, as the length of the input sequence grows, RNNs are prone to losing informa-
tion from the beginning of the sequences due to vanishing and exploding gradients issues

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Table 4  Variants of CNN with their application areas
Variants of CNN [Ref.] Features Application

CNN [61] It encodes the entire source sentence simultaneously Neural Machine Translation
Multichannel CNN [130] It makes several predictions using multi-task learning Question Identification,
Sentence Classification
Multimedia Tools and Applications (2023) 82:43089–43144

Dynamic CNN [39] It maps the meanings of words to a sentence Document Summarization
Multi-column CNN [44] It uses multiple column networks to extract information from the input context Question Answering
Dynamic multi-pool CNN [27] It incorporates a dynamic multi-pooling layer that employs event triggers and arguments to Event Extraction
retain critical data from the pooling layer
Hybrid CNN-RNN [87] Along with the frequency axis, it provides frequency shift-invariance Image Captioning

13
43105

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

43106 Multimedia Tools and Applications (2023) 82:43089–43144

Fig. 9  Recurrent Neural Network

[16, 146]. RNNs fail to model the long-range dependencies of natural languages. Conse-
quently, Long short-term memory [74] and gated recurrent unit [32] have been proposed as
alternative recurrent networks that are better prepared to learn long-distance dependencies.
These units are better at learning to memorize only the part that is relevant for the future.
At each time step, they dynamically update their states, deciding on what to memorize and
what to forget from the previous input. The LSTM cell has separate input and forget gates
as shown in Fig. 10, while the GRU cell performs both of these operations together using
its reset gate.
The forget gate decides which information of the long-term memory is useful and which
to forget. The next input gate determines which new information to be added to the net-
work, and the final output gate decides the new hidden state. In a vanilla RNN, the entire
cell state is updated with the current activation, whereas both LSTMs and GRUs have the
mechanism to keep the memory from previous activations. This allows recurrent networks
with LSTM or GRU cells to remember features for a long time and reduces the vanishing
gradient problems as the gradient back propagates through multiple bounded nonlineari-
ties. LSTMs and GRUs have been very successful in modelling natural languages in recent
years

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Table 5  Variants in RNN with their application areas
Variants of RNN [Ref.] Features Application

Simple RNN [29] It uses the idea of processing sequential information Machine translation, Sentence classifica-
tion, Question Answering, language
modelling
Bi-directional RNN (BRNN) [175, 187] It allows training the network in both time directions simultaneously Machine Translation, Speech Recognition
RNN-EM (RNN with External Memory)[152] The use of external memory improves the memorization capability of RNNs Language Modelling
Multimedia Tools and Applications (2023) 82:43089–43144

GF-RNN (Gated Feedback RNN) [31] It controls the strength of the temporal connection adaptively and allows multiple Character-level language modelling
adaptive timescales
p-RNN (Parallel-RNN) [73] It leverages the added value of multiple item representations Session-based recommendations
Q-RNN (Quasi RNN) [19] It allows for parallel computation, thus enabling high throughput and good scaling Language Modelling, Document-level
to long sequences sentiment classification, Character-level
machine translation
MV-RNN (Multi-View RNN) [36] It monitors changes in user’s preferences over time Sequential Recommendation
CRNN (Convolutional RNN)[174] It combines features of CNN and RNN Text Recognition

13
43107

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

43108 Multimedia Tools and Applications (2023) 82:43089–43144

Fig. 10  Long Short Term Memory (LSTM)

Fig. 11  Reinforcement Learning

4.2.4 SeqGANs and reinforcement learning

SeqGAN (Sequential Generative Adversarial Network) is a variant of GAN (Generative

Adversarial Network), used to generate text. This SeqGAN combines reinforcement learn-
ing and GANs for learning from discrete sequence data [231]. In SeqGANs, the generator
is treated as an RL agent. The tokens generated till a particular time become the state. The
token to be generated next is the action and the reward is the feedback given by the dis-
criminator to guide the generator in evaluating the generated sequence.
Reinforcement learning (RL) is a gradual stamping of behaviour [86, 166] where an
agent learns how to act in an environment by performing actions and analyzing the out-
comes, as shown in Fig. 11. The performance is maximized by allowing software agents
and machines to determine the ideal behaviour within a specific context automatically. The
agents are required to learn their behaviour using simple reward feedback, known as the
reinforcement signal.
When using reinforcement learning for automated text generation, the actions are writ-
ing words, and the states are the words already written by the algorithm. The actions,
rewards, and policies corresponding to text generation tasks are mentioned in Table 6.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43109

4.2.5 Transformer models

The Transformer models introduced by [195] have facilitated the enhancement of a wide
range of text generation tasks. Transformer models are based on global dependencies
between the input and output, use attention mechanisms, and have the capability to cap-
ture the linguistic knowledge of vocabulary, syntax, and semantics. The transformer is an
encoder-decoder architecture. RNNs and LSTM architectures have significant difficulties
with longer sequences as a result of the vanishing gradient problem [16, 146]. The prob-
ability of keeping context from a word that is further away from the word that is being pro-
cessed diminishes exponentially as the sentence grows longer. Parallelization is a practical
approach for training on larger datasets. The transformers expand with data and architec-
tural size, capture longer sequence features, and enables parallel training. As a result, more
effective and coherent language models are feasible. Prior to this, most of the ATG models
were trained on supervised learning. However, supervised models need a large amount of
annotated data for learning a particular task which is often not easily available, and they
fail to generalize for other tasks.
The Transformer is a sequence-to-sequence model and consists of an encoder and
decoder. Both encoder and decoder are multiple identical blocks layered on top of each
other. The overall architecture of the Transformer is shown in Fig. 12. Each encoder block
consists of a multi-head self-attention module followed by a position-wise feed-forward
network (FFN). Around each module, a residual connection is employed, followed by the
layer Normalization. Compared to the encoder blocks, decoder blocks additionally insert a
third module, known as encoder-decoder attention, between the multi-head self-attention
and FFN module. Furthermore, the masked attention module preserves the auto-regressive
property, ensuring to prevent each position from attending to subsequent positions.
Transformer models such as BERT [40], and GPT-3 [20] are pre-trained on large
corpora and use unsupervised learning for text generation. These pre-trained transformers are
classified into three categories, namely: encoder-only (like BERT), decoder-only (like GPT-
n), and encoder-decoder (like T5). Bidirectional Encoder Representations from Transformers
(BERT) is the first deep bidirectional, unsupervised language representation [40] model. It is
built upon work in contextual representations. BERT uses an attention mechanism along with
a Transformer that learns contextual relations in text to generate a text [40]. Generative Pre-
Trained Transformer (GPT) is an autoregressive language model that was trained with 175
billion parameters to generate text automatically. It uses unlabeled data and then fine-tuning
for the specific downstream task. GPT-n series (GPT-1 [162], GPT-2 [20], GPT-3 [20]) shows
significant performance on various ATG tasks even without finetuning or gradient updates.
Transformer-based models, such as GPT, T5, XLNet, and BERT [20, 40, 163], showed
impressive results on several text generation tasks such as question answering, language
modelling, machine translation, sentiment analysis, and summarization, as shown in Table 7.

Discussion on research question 3

With the advancements in deep neural networks, text generation models are capable of
generating realistic, fluent, and coherent natural language. This section helped to answer
RQ 3- “What are the various approaches and associated architectures in the field of text
generation.” It has been observed that the field of text generation has undergone significant

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 43110

13
Table 6  Policy, Action and Reward function for different Applications
Application Policy Action Reward

Text Summarization, Question Attention-based models, pointer- Selecting the next word for a headline, translation, and summary ROUGE, BLEU
Generation generators,
Question Answering Sequence-to-sequence model Selecting the answer from input document or selecting the start and end F1 score
index of the answer
Image/Video Captioning Sequence-to-sequence model Selecting the next token for the caption CIDEr, SPICE, METEOR
Dialog Generation Sequence-to-sequence model Dialogue utterance to generate BLEU, Length, and
Diversity of dialogue
Multimedia Tools and Applications (2023) 82:43089–43144

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Multimedia Tools and Applications (2023) 82:43089–43144 43111

Fig. 12  Transformer architecture

changes from template-based and statistical-based traditional approaches to, most recently,
pre-trained transformer based models. As a result of these advancements in techniques,
the text generation research field has witnessed remarkable progress and a surge in interest
for the study. The shift starts with recurrent neural networks, long short-term memory
networks, gated recurrent units for learning language representations, and later sequence-
to-sequence learning, which opens a new chapter characterized by the wide application
of the encoder-decoder architecture. However, these sequence-to-sequence models cannot
capture long term dependencies, which motivated the development of pointer networks
and attention networks. Then, the transformer architecture incorporates an encoder and a
decoder with self-attention mechanism, which is now widely used by text generation tasks.
The availability of powerful deep neural models, computationally intensive architecture
and pre-trained transformer models results in the incredible adoption of a variety of text
generation applications, including text summarization, machine translation, creative

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 43112

13
Table 7  Transformer Models with parameters and datasets
Transformer Model Variations with parameters Dataset used Application

BERT (Bidirectional Encoder Represen- BERT_Base (110 million parameters), 16 GB (Books Corpus + Wikipedia) + 3.3 Question Answering, Text Summarization,
tations from Transformers) [40] BERT_Large (345 million parameters) Billion Words Machine Translation
DistilBERT (distilled version of BERT) Base (66 million parameters) 16 GB (Books Corpus + Wikipedia) + 3.3 Question Answering, Language Modelling
Billion Words
RoBERTa (Robustly Optimized BERT Base (110 million parameters), Large 16 GB (Books Corpus, Wikipe- Language Modelling
Pretraining Approach) (345 million parameters) dia) + 144 GB (38 GB Web text corpus,
76 GB of Common Crawl News
dataset)
XLNet (combines the bidirectional capa- Base (110 million parameters), Large 130 GB (Books Corpus + Wikipe- Question Answering, Sentiment analysis,
bility of BERT with the autoregressive (345 million parameters) dia + additional) + 33 Billion Words Document Ranking
technology of Transformer-X)
GPT-1 (Generative Pre-trained Trans- 117 M parameters BooksCorpus dataset NER, Language Modelling
former) [162]
GPT-2 [20] 1.5 billion parameters WebText corpus (8 million documents), Comprehension reading, Summarization,
Children’s Book dataset, LAMBADA Translation
dataset
GPT-3 [20] 175 billion parameters Common Crawl, WebText2, Books1, Machine translation, transfer learning
Books2, Wikipedia
T5 (Text-to-Text Transfer Transformer) T5-Small (60 million param- C4 (Colossal Clean Crawled Corpus) Summarization, Question Answering,
[163] eters), T5-Base (220 million Machine Translation, News Article
parameters), T5-Large (770 million generation
parameters), T5-3B (3 billion param-
eters), T5-11B (11 billion parameters)
Multimedia Tools and Applications (2023) 82:43089–43144

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Multimedia Tools and Applications (2023) 82:43089–43144 43113

applications such as story generation, and dialogue generation. However, applying these
models to different neural text generation tasks can depend on various factors, including
the type of task, the architecture of the model, the size and quality of the training data, and
the evaluation metrics used to measure the performance of the model.
The deep learning approaches arise due to the availability of a large number of corpora
and significant computational resources. The standard datasets for text generation applica-
tions are mentioned in the next section.

5 Datasets for text generation tasks

In research, the datasets have been used to assess the performance of a proposed method.
The deep learning models trained on large-scale datasets demonstrate unrivalled abilities
to understand patterns in the data, opening a whole slew of new possibilities for creating
realistic and coherent texts. Several datasets were recently created to support the training of
text generation models. The datasets vary in terms of output lengths, generation tasks, and
domain specificity. This section describes some of the datasets that are commonly used in
text generation tasks. In Table 8, a shortlist of some of the task-specific standard datasets is
provided, which is organized by the text generation applications.
Each dataset has many files, including training, testing, and validation files, in various
formats. Few datasets have files in json format, text format, and excel format, while others
are in csv format. For a better understanding of the above-mentioned datasets, screenshots
of few datasets have been provided in the Fig. 13 below.

Discussion on research question 4

The availability of large and diverse datasets has also benefited the recent progress
in text generation. This section helped to answer RQ 4- “What are the available datasets
adopted in which the stated applications are organized.” It has been observed that there
are many datasets that vary in terms of text generation tasks and domain specificity. The
datasets help to train, test and validate the text generation models. Nowadays, there is a
trend to train models on massive datasets. However, training text generation models on
diverse datasets provide the opportunity to improve their robustness. The models trained on
massive datasets show an unmatched ability to automate the generation of fluent and coher-
ent texts. Thus, while training a text generation model for a particular task, it is critical to
choose the dataset carefully. For a specific text generation task, the most commonly used
datasets are shown in Fig. 14. It also specifies the percentage of articles in which a given
dataset is used for a specific text generation task.
After training the model, text decoding plays a vital role in the generation of text. The
following sections discuss text decoding techniques and optimization methods for text
generation.

6 Text decoding techniques and optimization algorithms

The automatic text generation model aims to generate text that is as good as human-written
text. And after training the model, the quality of the generated text has a significant impact
on the decoding strategy and optimization technique that one employs. In this section,

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Table 8  Application-specific Text Generation Datasets
43114

Application Dataset [Year] Description URL

13
Text Summarization CNN-Daily Mail dataset [2017] Around 287 K news articles, each with https://​github.​com/​abisee/​cnn-​daily​mail
upto four summaries
Text Summarization Newsroom dataset [2018] 1.3 M news articles and metadata https://​summa​ri.​es/
Text Summarization DUC-2003 and DUC-2004 (Document 500 news articles, each with four dif- https://​duc.​nist.​gov/​data.​html
Summarization Challenge) [2003, ferent human-generated reference
2004] summaries
News Summary XSum [2018] One sentence summary of the input https://​github.​com/​shash​iongi​thub/​XSum
news article, 3,99,147 documents with
81,092 summaries
Headline Generation Gigaword [2015] It contains 8 M + news articles https://​github.​com/​harva​rdnlp/​sent-​summa​
ry
News Summary NY Times articles [2008] 1,399,358 documents, 294,011 sum- https://​www.​kaggle.​com/​nzala​ke52/​new-​
maries york-​times-​artic​les?​select=​nytim​es_​
news_​artic​les.​txt
Question Answering and Question Stanford Question Answering data set 100 k + question–answer pairs collected https://​rajpu​rkar.​github.​io/​SQuAD-​explo​
Generation (SQuAD) (1.0 and 2.0) [2016, 2018] by crowdsourcing through compilation rer/
of Wikipedia articles
Question Answering and Question TriviaQA [2017] 650 k question–answer-and-evidence http://​nlp.​cs.​washi​ngton.​edu/​trivi​aqa/
Generation triples
Yes/No Questions BoolQ [2019] 16 k Questions with answers https://​github.​com/​google-​resea​rch-​datas​
ets/​boole​an-​quest​ions
Multiple Choice Questions ARC Corpus [2018] Around 8 k science questions with multi- https://​allen​ai.​org/​data/​arc
ple choice answers
Question Answering QuAC [2018] 100 k QA pairs https://​quac.​ai/
Question Answering HotPotQA [2018] 113 k Wikipedia-based question–answer https://​hotpo​tqa.​github.​io/
pairs
Reading Comprehension with Common- ReCoRD [2019] 70,000 + news articles with https://​sheng-z.​github.​io/​ReCoRD-​explo​
sense Reasoning 1,20,000 + queries, each query question rer/
has answer from corresponding news
Multimedia Tools and Applications (2023) 82:43089–43144

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Table 8  (continued)
Application Dataset [Year] Description URL

Conversational Question Answering CoQA dataset [2019] 127 k conversational questions with free- https://​stanf​ordnlp.​github.​io/​coqa/
form text answers
Dialogue Generation OpenSubtitles [2016] It contains conversations for 20 k + mov- http://​opus.​nlpl.​eu/​OpenS​ubtit​les.​php
ies between movie characters
Dialogue Generation Cornell Movie Dialogues Corpus [2011] 20 k dialogues between more than 10 k https://​www.​cs.​corne​ll.​edu/​~crist​ian/​Corne​
movie characters ll_​Movie-​Dialo​gs_​Corpus.​html
Machine Translation WMT’ 14 [2014] 850 M words from English–French http://​www.​statmt.​org/​wmt14/​trans​lation-​
parallel corpora of UN (421 M words), task.​html
Europarl (61 M words), news commen-
tary (5.5 M words), and two crawled
corpora of 90 M and 272.5 M words
Story Telling ROCStories [2016] Around 100 k five-sentence common- https://​www.​cs.​roche​ster.​edu/​nlp/​rocst​
sense stories ories/
Paraphrase Generation Microsoft Paraphrase corpus [2005] 5800 pairs of semantic equivalent sen- https://​www.​micro​soft.​com/​en-​us/​downl​
Multimedia Tools and Applications (2023) 82:43089–43144

tences taken from online news sources oad/​detai​ls.​aspx?​id=​52398

with human annotations
Paraphrase Generation ACL’s Semantic Text Similarity competi- It includes text from news headlines, http://​ixa2.​si.​ehu.​es/​stswi​ki/​index.​php/​
tion [2017] image captions, and user forums for STSbe​nchma​rk
2012–2017
Paraphrase Generation Quora Duplicate Questions [2017] Around 400 k question pairs (of these http://​qim.​fs.​quora​cdn.​net/​quora_​dupli​
around 111 k questions occur across cate_​quest​ions.​tsv
multiple pairs)
Image Captioning COCO data set [2021] 330 k images http://​cocod​ataset.​org/
Video Captioning MSR-VTT [2017] 10 k videos and each containing 20 http://​ms-​multi​media-​chall​enge.​com/​2017/​
human-annotated captions chall​enge
Video Captioning YouTube2Text/MSVD [2011] 1970 videos and each 40 human-anno- http://​www.​cs.​utexas.​edu/​users/​ml/​clamp/​
tated captions video​Descr​iption/

13
43115

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

43116 Multimedia Tools and Applications (2023) 82:43089–43144

Fig. 13  Screenshots of datasets [a CNN/DM dataset for summarization task b Gigaword dataset for sum-
marization task c SQuAD dataset for question-answering task d Empathetic dialogues dataset for dialogue
generation task e ROC story dataset for story generation and question-answering task f WMT dataset for
translation task g Quora question pairs dataset for paraphrasing task h Flickr30k dataset for captioning task]

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43117

Text Summarization Dataset Summary

Xsum Machine Translation Dataset Summary
Summary from GitHub
SNLI WMT

Dataset names
Radiology report Private dataset
Dataset names

NYT NIST
Newsela Newstest
MSR-ATC ILCI Corpus
LCSTS HindiEnCorp
Gigaword BTEC
DUC
Debatepedia 0 10 20 30 40 50
CNN/DM
%age of papers in which the dataset is used
0 20 40
%age of papers in which the dataset is used d: Summary of machine translation datasets

a: Summary of text summarization datasets Story Generation Dataset Summary

Writing Prompts
Question Answering Dataset Summary
Wordnet

Dataset name
Wikipedia Movie Plots corpus
Wikimovies
VIST
WebQuestions
TriviaQA TV show plot summaries
SQuAD Story Cloze
Slides from slidewiki ROCStories
SimpleQuestions private dataset
SciQ English-language stories
QuAC
QGSTEC dataset 0 20 40
Dataset name

QALD %age of papers in which the dataset is used

private dataset
Penn Treebank (PTB)
NewsQA
e: Summary of story generation datasets
NarrativeQA
MCQL Paraphrase Generation Dataset Summary
IMDB movie reviews
ILSVRC-CLS
WikiAnswers
Free 917
Dataset name

CBQ corpus Twitter URL paraphrasing…

Campbell's Biology textbook Quora question pairs
C4
MSRP
BoolQ
AQAD (Amazon QA corpus) MS COCO
30M Factoid Question… ML-Paraphrase
0 10 20 PPDB
%age of papers in which the dataset is used 0 10 20 30

%age of papers in which the dataset is used

b: Summary of question answering datasets
f: Summary of paraphrasing datasets
Dialogue Generation Dataset Summary
Twitter Conversation Dialogues Image/Video Captioning Dataset Summary
PERSONA-CHAT
TV Series Transcription PASCAL
Dataset name

STC MS COCO
Dataset names

Public Transport Information Instagram

Open Subtitles ImageNet
NLPCC IAPR
Movie Triples Flickr8K/30K
IT Help Desk Abstract Scenes
Empathetic Dialogues 0 10 20 30 40 50
Cornell Movie Dialogues
%age of papers in which the dataset is used
Baidu Teiba Forum

0 10 20 g: Summary of image/video description datasets

%age of papers in which the dataset is used

c: Summary of dialogue generation datasets

Fig. 14  Summarized percentage of datasets usage in particular text generation task [a Summary of text
summarization datasets b Summary of question answering datasets c Summary of dialogue generation
datasets d Summary of machine translation datasets e Summary of story generation datasets f Summary of
paraphrasing datasets g Summary of image/video description datasets]

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43118 Multimedia Tools and Applications (2023) 82:43089–43144

Fig. 15  Autoregressive method

decoding methods and optimization techniques important for text generation have been
discussed.

6.1 Decoding techniques

Decoding is the process of generating natural language texts from a model. As there is a
need for one-to-one correspondence between input and output time steps of generation,
which leads to a crucial key aspect named decoding. The decoding approach in a neu-
ral text generation system describes how the system searches for potential output utter-
ances when generating a sequence. It specifies how the words are combined to form text
and sentences. Without an appropriate decoding technique, the generated text results in
vague and dull text.
Primarily, decoding can be categorized as autoregressive and non-autoregressive [215].
In an autoregressive generation, the target tokens are generated one by one in a sequential
manner, as shown in Fig. 15. The beginning and end of decoding are controlled by special
tokens, including [BOS] (beginning of a sentence) and [EOS] (end of a sentence), which
implicitly determine the target length during decoding.
Traditional models capture the true distributions of words using this strategy. The fun-
damental reason is the conditional dependence property from left to right. The transformer-
based models that cannot replicate the training benefits as training can be nonsequential,
and inference holds to be sequential with autoregressive decoding is among one issue with
this technique. Another issue with the autoregressive approach is that it is time-consuming,

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43119

Fig. 16  Non-autoregressive
method

especially for generating long target sentences. To alleviate this problem and accelerate
decoding, non-autoregressive generation is proposed [65, 72]. The non-autoregressive
decoding technique can generate all the target tokens in parallel, as shown in Fig. 16.
In this method, as all the target tokens are generated in parallel, there is no need for spe-
cial token or target information to guide the termination of decoding. Using this technique,
inference speed is hugely increased. [115, 157] proposes non-autoregressive techniques
for summarization. The autoregressive techniques can be further viewed as sampling and
search techniques. There are many different decoding strategies, including greedy, beam
search, random sampling, top-k sampling, and nucleus sampling, as discussed.
A greedy search selects the most probability word from the language model. It uses
this word as the next word and feeds it as input on the next step till it reaches maximum
length [207]. However, greedy search is bound in a loop of the same words, resolved by
random sampling [14]. Greedy search also lacks backtracking, which results in unnatural
and meaningless sentences. The greedy search is not optimal for generating high-proba-
bility sentences [109], and this problem has been addressed by the beam search decoding
method.
Random sampling picks the word randomly according to the conditional word probabil-
ity extracted from the text generation model [14]. However, directly using the probabilities
extracted from the text generation models often leads to incoherent text. Moreover, this text
decoding method is not deterministic. Nevertheless, applying a softmax over the probabil-
ity distribution and varying its parameter makes it smoother.
Beam search keeps the most probable words by tracking multiple possible sequences at
once. It keeps track of the k-most probable partial sequences at each step, where k is the
beam size. Beam search chooses the words to obtain an overall highest probability sentence
[109, 198]. The text generated with beam search is more fluent as compared to greedy

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43120 Multimedia Tools and Applications (2023) 82:43089–43144

search. But when beam size is equal to one, beam search behaves as greedy search. Beam
search produces a list of nearly identical sequences that fail to capture the inherent ambi-
guity of complex text generation tasks. Diverse Beam Search overcomes this problem by
describing beam search as an optimization problem and augments the objective with diver-
sity [197].
In top-k sampling, the ‘k’ most likely next words are selected, and then the next pre-
dicted word is sampled only from these ‘k’ words [53]. As’k’ is fixed in top-k sampling,
the number of words filtered from the next word probability distribution is not dynamically
modified [205]. As a result, unlikely words may be selected among these ‘k’ words if the
next word probability distribution is very sharp.
Nucleus sampling (or top-p sampling) selects words from the smallest possible set with
a cumulative probability greater than some probability p. As a consequence, the number of
words in the set can dynamically decrease and increase according to the next word prob-
ability distribution [75]. It is the best available decoding strategy for generating long-form
text that is both high-quality as measured by human evaluation and as diverse as human-
written text.
Other decoding techniques include semi-autoregressive, iterative, and mixed decoding
[215]. The semi-autoregressive decoding generates multiple target tokens at one decod-
ing step. The iterative decoding provides target information on each decoding step. Some
works aim to combine these decoding strategies into a unified model. Tian et al. [190] pro-
pose a unified approach for machine translation that supports autoregressive, iterative, and
autoregressive decoding methods.
Thus, based on the strengths and limitations of the text decoding techniques, the choice
of decoding method has a significant impact on the linguistic features and the quality of the
generated text.

6.2 Optimization techniques

With the tremendous growth in the amount of data, optimization has become an essen-
tial part of deep learning. The goal of the optimization algorithm is to minimize the loss
function by reaching global semi-minima. Deep Learning models are becoming efficient
and achieve better results with the use of optimization techniques. This section describes
the commonly used optimization methods from a neural text generation perspective.
There has been much interest in modifying the stochastic gradient descent algorithm
with an adaptive learning rate for more stable training, e.g., AdaGrad, AdaDelta, and
Adam, as shown in Fig. 17.
Further, these optimization algorithms are reviewed in a summarized manner based
on their properties, pros, and cons, as mentioned in Table 9.

Discussion on research question 5

The text decoding methods and optimization techniques significantly impact the quality
of the generated text This section helps to answer RQ 5- “What are various text decod-
ing and optimization techniques used to automatically generate text.” These techniques
can be applied to different text generation tasks, i.e., text summarization, story generation,
paraphrasing task, translation, and image captioning. Based on the text generation task,
the autoregressive and non-autoregressive decoding technique can be utilized. Using a
good model with bad decoding strategies lead to repetitive loop problem and inconsistent,

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43121

Fig. 17  Optimization algorithms

for Text Generation

incoherent text generation problems. As a result, it is recommended to choose a decoding strat-

egy carefully. Decoding methods like, top-k sampling and nucleus sampling produce more flu-
ent text than beam search and greedy search. However, top-k sampling has suffered from gen-
erating repetitive word sequences recently. There has also been observed that greedy and beam
search perform better if a different training objective is used by the text generation model.
In an end-to-end neural framework, all kinds of inputs, including target generated text
are firstly mapped into numeric embeddings, and then neural modules feed-forward infor-
mation layer by layer. Finally, the last output of the neural framework is used to generate
the target tokens with a decoding strategy and calculate the losses to optimize parameters.
Optimization algorithms are among those parameters and play an important part to infer
the losses between neural networks. Thus, optimization algorithms are equally important
for text generation models. It has been observed that the training performance of the model
is influenced by the selection of an optimization algorithm. After understanding the con-
cepts of various optimization algorithms and the function of their parameters, text genera-
tion models perform better. For a particular text generation task, the most commonly used
optimizers are shown in Fig. 18. It also specifies the percentage of articles in which a given
optimizer is used for a specific text generation task. There are many real-time text genera-
tion tools which are discussed in next section.

7 Real‑time text generation tools

Text generation has played an essential role in various applications of text generation, such as paraphras-
ing, question generation, summarization, and dialogue systems. Text generation systems assist human
writers and make the writing process more effective and time-saving. This section describes several
real-time tools for text generation. The tools for text generation applications are mentioned in Table 10.

Discussion on research question 6

The existing automatic text generation application tools have been able to generate inter-
esting text but are limited in terms of consistency, fluency, controllability, and diversity of

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 43122

13
Table 9  Summarized review of Optimization algorithms
Method [Ref.] Properties Pros Cons

Gradient Descent It minimizes a given function, computes It is locally optimal and globally con- It performs redundant computations for
the gradient of the cost function vergent large datasets
Stochastic Gradient Descent (SGD) An iterative method that optimizes an It is simple and fast convergence for tasks It is implicitly sequential, parallelizing it
objective function having an extensive training set with GPUs is difficult
AdaGrad [49] Adapts the learning rate to the parame- Infrequent parameters have larger The learning rate shrinks as the accu-
ters, shrinks the learning rate adaptively updates, and periodic parameters have mulated sum grows, gradually being
smaller updates infinitely small
AdaDelta [234] Adapts learning rates based on a moving Even after several updates, it continues Not suitable for stationary problems
window of gradient updates learning
Root-Mean-Square propagation It collects the gradient for the learn- It performs well in a non-convex setting The upgrade process may be replicated
(RMSprop) ing rate as an exponentially weighted with differences between global and around the local minimum in the late
average local structures training stage
Adaptive Moments (Adam) [89] It combines several properties from It has adaptive learning rates for each The method may not converge in some
Momentum, RMSprop, AdaGrad parameter cases
Nesterov Momentum to Adam (Nadam) It combines NAG and ADAM Achieves best results with Nesterov Not always the best algorithm to choose
[45] momentum
SGD with momentum [158] It makes use of past gradients to update It accelerates SGD in the relevant direc- It makes use of only simple momentum
the trainable parameters of the neural tion, reduces oscillations, and faster
network convergence
Nesterov Accelerated Gradient Descent It evaluates the gradient with the momen- Momentum helps to jump out of locally It is challenging to choose a suitable learn-
(NAG) [45] tum optimal solutions ing rate
Multimedia Tools and Applications (2023) 82:43089–43144

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

Multimedia Tools and Applications (2023) 82:43089–43144 43123

Text Summarization Optimizer Machine Translation Optimi zer

Summary Summary
RMSProp
3% Adadelta
22%
Adagrad Adam Softmax
7% 39% 11%

SGD
24%

Adadelta SGD
27% 67%

a: Summary of text summarization Optimizer d: Summary of Machine translation Optimizer

Question Answering Optimizer Story Generation Optimizer Summary

Summary

RMSProp Adam Softmax Adadelta

6% 26% 7% 11%
Adam
30%
Adagrad
SGD
9%
23% SGD
33%
Cross
Adadelta
AdaFactor Entropy Adamax
21%
15% 15% 4%
b: Summary of Question Answering Optimizer e: Summary of story generation Optimizer

Dialogue Generation Optimizer Paraphrase Generation Optimizer

Summary Summary
Softmax
5% Adagrad
SGD Softmax
Adadelta 19% 6% 17%
9%

Cross
Adagrad Entropy Adam
5% 19% 37%
SGD
37%
Adam
38% Adamax Adamax
5% 3%
c: Summary of Dialogue Generation Optimizer f: Summary of paraphrasing Optimizer

Image/Video Captioning Optimizer

Summary
Softmax Cross
7% SGD Entropy
21% 5%
Adadelta
14%

Adagrad
10%
Adamax
Adam
17%
26%

g: Summary of image/video description Optimizer

Fig. 18  Summarized percentage of optimizers used in particular text generation task [a Summary of text
summarization Optimizer b Summary of Question Answering Optimizer c Summary of Dialogue Genera-
tion Optimizer d Summary of Machine translation Optimizer e Summary of story generation Optimizer f
Summary of paraphrasing Optimizer g Summary of image/video description Optimizer]

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Table 10  Real-time text generation application tools
43124

Tool Name Application area Features Access Link

13
Sassbook AI Summarizer Text Summarizer It understands the context and generates summaries in its https://​sassb​ook.​com/​ai-​summa​rizer
own words, which captures the essence of the topic
Summarize Bot Summary Generator It extracts and summarizes information from weblinks, https://​www.​summa​rizeb​ot.​com/
news articles, scientific articles, books, e-mails, lectures,
patents, and legal documents
Resoomer Summary Generator It summarizes argumentative texts and articles, scientific https://​resoo​mer.​com/
texts, and educational documents
Automatic Summarizer Text Summarizer It generates multi-language summarization https://​autos​ummar​izer.​com/
SMMRY Summary Generator It creates a summary of articles and text in the specified https://​smmry.​com/
number of lines
Text Compactor Summary Generator It is a free online summarization tool mainly for textbooks https://​www.​textc​ompac​tor.​com/
Flexudy Question generation, Summary generator It creates WH-question, fill-in-the-blanks, and summaries https://​flexu​dy.​com/
out of user text
Lumos Learning Question–Answer generator, Summarization It provides automated assessment questions and summari- https://​www.​lumos​learn​ing.​com/​llwp/
zation of test reports
GeneratorQ Question–Answer generator It generates a variety of questions and even answers based https://​gener​atorq.​azure​websi​tes.​net/
on inputted text
Questo Question generation It generates questions from text and images of text as well https://​questo.​ai/
Quillionz Question/Answer Generation It generates a wide range of questions, including multiple- https://​www.​quill​ionz.​com/
choice
RevUp Question Generation It generates quality tests for self-motivated learners and https://​paper​swith​code.​com/​paper/​
teachers revup-​autom​atic-​gap-​fill-​quest​ion-​
gener​ation
AceBot Conversation generator It guides survey respondents to complete a survey through https://​acebot.​ai/
a fun and interactive conversation through chat messages
Pana Conversational Agent It is a virtual travel agent that suggests places to visit and -
where to eat
Sephora Conversational Agent It assists in the selection and purchase of cosmetics as well -
as the acquisition of beauty advice
Multimedia Tools and Applications (2023) 82:43089–43144

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

 Table 10  (continued)
Tool Name Application area Features Access Link

HealthTap Conversational agent It allows people to ask questions, get test results evaluated, -
and receive feedback from physicians, making health-
care accessible
DeepL Translation It understands and translates text and helps in overcoming https://​www.​deepl.​com/​home
language barriers
Anusaaraka Translation It does the translation of English text to Hindi text http://​sampa​rk.​iiit.​ac.​in/​anusa​araka/
Plot Generator Story Generation It creates customized short stories https://​www.​plot-​gener​ator.​org.​uk
Storyline Creator Story-telling It allows for drafting ideas and organizing scenes and https://​www.​story​linec​reator.​com/
characters, allowing users to focus on the story plot and
the creativity
Creative Help Story Generation It creates stories where users interact with the generated http://​www.​get-​creat​ive.​help
text
Converse Smartly Speech to text It automatically transcribes audio from the English https://​www.​folio3.​ai/​conve​rse-​smart​ly/
language in real-time, and has been detecting multiple
Multimedia Tools and Applications (2023) 82:43089–43144

speakers as well
QuillBot Paraphrasing tool It helps in rewriting and enhancing any sentence, para- https://​www.​quill​bot.​com/
graph, or article
Paraphrase Online Paraphrasing tool It helps in automatic online paraphrasing of the article https://​www.​parap​hrase-​online.​com/
Wide Eyes Image Captioning It saves time and effort by generating tags for objects in https://​widee​yes.​ai/
the images for the fashion and e-commerce industry
Imagga Image Captioning It identifies the content of visuals, analyzes, and extracts https://​imagga.​com/
their features

13
43125

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

43126 Multimedia Tools and Applications (2023) 82:43089–43144

the generated text. This section helps to answer RQ 6- “What real-time tools are available
for automatic text generation tasks.” It has been observed that there are diverse categories
of automated text generation tools for different applications in the real world, such as some
excel in generating short texts like headlines or tweets, and others excel at generating long
texts like articles or blog entries. However, the authenticity of the content is missing with
automated text generation tools, thus the fake/inaccurate content is roaming around. Many
sectors have started using automated text to improve user experience as the algorithms are
capable of generating human-like texts. But the automated tools can be exploited nega-
tively. These tools can be abused by students who want to cheat on school work and ham-
pers the student’s ability. The content generated by these tools is fast and cheap but lacks
the artistry involved in expressing thoughts. However, some of the observed tools are aver-
age in text generation, while others generate fluent text but are not freely accessible. Some-
times the text generated by these tools is superficial and repetitive. Thus, there is still much
research being done and many problems to be solved, including long-term dependencies,
redundancy while generating text, word sense ambiguity, incorrect grammar, consistency,
and many more. These tools are limited by the data they were trained on and may not have
a deep understanding of the topic one is writing about. The automated text generators can-
not provide original and creative ideas, and makes people lazy and dependent on automa-
tion. For the effective text generation and the reliable assessment of the text generation
models, there are many task-specific evaluation measures, as described in the next section.

8 Evaluation metrics for text generation

This section discusses the automatic evaluation measures that are frequently used to assess
the advancements in the text generation system. Without proper evaluation, it is difficult to
measure a system’s competitiveness, which hinders the development of advanced algorithms
for text generation. The goal of evaluation metrics is to evaluate the effectiveness of text gen-
eration tasks, and for this, a robust and unbiased evaluation metric is important. An automatic
metric that correlates well with human assessments is ideal. It is desirable to employ a variety
of metrics to assess the efficiency of the system over multiple aspects. The most popular auto-
mated evaluation methods for evaluating machine-generated text are mentioned in Table 11,
with the pros and cons of the metric.
Automated text evaluation metrics are used to assess the text generation models, such
as question–answer generation, text summarization, or machine translation. These evalu-
ation measures provide a score that reflects the similarity between a human written ref-
erence text and an automatically generated text. There are many criteria based on which
one decides which metric to use for which text generation task, as shown in Table 12.

Discussion on research question 7

As the field of text generation is continually advancing, evaluation is becoming criti-

cal for assessing progress in the area and performing comparisons between text generation
models. This section answered the RQ 7- “Which metrics or indicators are used to evalu-
ate automated text generation.” It has been observed that traditionally language models
have been evaluated based on perplexity, which concerns with the probability of a sentence
being produced by the model. There are many well established automated evaluation met-
rics for assessing specific text generation tasks, such as METEOR and ROUGE for text

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
 Table 11  Summarized review of automatic evaluation metrics for text generation
Metric [Ref.] Description Pros Cons

BLEU [143] It measures the similarity between two sentences It is simple to use It does not correlate well with human judgment
NIST [156] It evaluates the quality of the text Less frequent n-grams have heavy weights It uses heavier weights for rarer words
ROUGE [114] It evaluates the adequacy of the output text It includes median or mean score from individual It does not provide information of the grammar,
generated output text and the narrative flow of the generated text
METEOR [96] It counts the overlap of word or word units At sentence level, it yields a good correlation It treats all words equally in the reference set
between candidate sentence and reference with human judgments
sentence
CIDEr [196] It is a consensus-based protocol that analyzes a It reflects a high agreement with the human It sometimes causes unimportant details of a sen-
sentence’s similarity to a set of sentences in consensus tence to be weighted more
human form
WER [25] It calculates the edit distance between the hypoth- The metric is computed efficiently and is repro- There is a dependency on the reference sentences
esis sentence and references ducible
Multimedia Tools and Applications (2023) 82:43089–43144

SPICE [8] It employs scene graphs to calculate the correla- It reflects a strong correlation to human assess- It neglects the fluency of the captions generated
tion between reference and hypothesis text ments
RIBES [210] It relies on the order of words in the generated It is independent of word boundaries It shows a lower correlation to human assessment
text
WMD [133] It calculates the distance between two sequences It is hyperparameter free and performs well in As the length of the document becomes large, the
represented by relative word frequencies information retrieval relation between sentences is lost
SMD [33] It utilizes sentence embeddings to evaluate text in It correlates well with human evaluation It considers equal weight for each sentence in a
a continuous space document
TER [210] It computes the number of edits to the best cor- It assigns equal cost to all edits It considers only identical matches between the
responding reference reference and the hypothesis
BERTScore [236] It computes the token similarity of two sentences At system-level and sentence-level evaluations, it It computes only token similarity and not an exact
dousing contextual embeddings correlates well with human judgments match
HUSE [70] It measures both the diversity and the quality of It detects the low-diverse generations that For training the model, it relies on human judg-
the text generated humans fail to detect ments

13
43127

Content courtesy of Springer Nature, terms of use apply. Rights reserved.

43128 Multimedia Tools and Applications (2023) 82:43089–43144

Table 12  Summary of evaluation metrics with their application usage

Criteria Evaluation Metric SUM QAG DG MT SG PRG IC

n-gram based metrics BLEU ✓ ✓ ✓ ✓ ✓ ✓ ✓

NIST - - ✓ ✓ - - -
ROUGE ✓ ✓ - ✓ ✓ - ✓
METEOR ✓ ✓ ✓ ✓ ✓ ✓ ✓
CIDEr ✓ - - - - - ✓
WER - - ✓ - - - ✓
SPICE - - - - - - ✓
RIBES - - ✓ - - -
F-Score ✓ ✓ ✓ ✓ ✓ ✓ ✓
Distance-based Metrics TER - - - - ✓ ✓ -
EED ✓ - - - - - -
WER - - - ✓ - - -
Diversity-based Metric TTR​ - - ✓ - - - -
SELF-BLEU - - - - - - ✓
Distinct-k - - ✓ - - - -
Embedding-based Metrics WMD - - - - - - ✓
SMD ✓ - - ✓ - - -
RUBER - - ✓ - - - -
Learned Evaluation Metrics BERTScore ✓ ✓ - ✓ - ✓ ✓
MoverScore ✓ - - - ✓ - -
HUSE - - ✓ - ✓ - -
Human-Centric Evaluation Human Judgement ✓ ✓ ✓ ✓ ✓ ✓ ✓

where MT- Machine translation; QAG- Question–Answer Generation; SUM- Summarization; SG- Story
Generation; DG- Dialog Response Generation; PRG- Paraphrase generation; IC- Image Captioning

summarization, BLEU for machine translation, SPICE and CIDEr for image captioning.
However, there is no universal metric that suits all text generation tasks and reflects all
relevant features of text. These work well to judge the quality of the generated text that
the model has generated natural, human-like and grammatically correct sentences. How-
ever, with open-ended generation tasks such as story telling or dialogue generation, the
model is expected to not only produce high quality text but also to be creative and diverse.
Another important aspect to open-ended text generation is commonsense reasoning, which
is referred as consistency. Since the models are expected to produce much longer text, they
are more prone to generating illogical or factually incorrect sentences. BEAMetrics [176],
a Benchmark to Evaluate Automatic Metrics help in better understanding the strengths and
limitations of current metrics across a broad spectrum of tasks. Fast and reliable evaluation
metrics are key to progress in research. While traditional natural language generation met-
rics are fast, they are not very reliable. Conversely, new metrics based on large pretrained
language models are much more reliable, but require significant computational resources
[83]. It is important that language models are evaluated in all dimensions of open-ended
text generation—quality, diversity and consistency [140]. When evaluating language mod-
els on open-ended text generation task, it has been observed that Corpus-BLEU is the best

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43129

metric to evaluate the quality of generated text due to its similarity with human judge-
ment. As for diversity, Self-BLEU appears to be the best metric to use due to its simplicity
to calculate. To evaluate the consistency of the generated text, using selection accuracy
on the MultiNLI dataset is good enough for most cases. For specific task such as story
generation, other dataset can be considered such as StoryCloze. Thus, the evaluation of
large language models such as GPT is task specific [24, 63, 119]. While evaluating the
text generation models for efficiency, it is necessary to rely on multiple metrics that reflect
different text attributes such as fluency, grammaticality, coherence, readability, diversity,
etc., Though human evaluations represent the gold standard for assessing the quality of
machine-generated texts, but it is costly and time-consuming. As a result, automated meas-
ures for evaluation are to be used. But these automated evaluation measures should only
be used as a supplement to human judgments and not as a replacement. Also, the auto-
mated metrices of evaluation are to be used when they present a reasonable correlation
with human judgments.
However, there are still many issues or open problems in the generation of automated
text which are discussed in the next section.

9 Challenges of automated text generation tasks

Generating fluent, meaningful, well-structured, and coherent text is pivotal for many text
generation tasks. It takes significant effort by humans to model long-term dependencies
while generating consistent text. It is an equally challenging task to do it automatically due
to the discrete nature of textual data. This section identifies the main difficulties or chal-
lenges for the effective generation of text as below.

9.1 Text summarization

Text summarization is a challenging task since it requires thorough text analysis to pro-
vide a reliable summary [57]. A good summary must include relevant details and must be
precise, but it must also consider aspects such as non-redundancy, significance, coverage,
coherence, and readability [145]. To achieve all these things, in summary, is a major chal-
lenge. While many text summarization models provide tangible results, several issues are
being suppressed. They often tend to repeat factually inaccurate information, struggle with
Out of Vocabulary (OOV) words, emphasize a word/phrase several times, and are also a bit
repetitive [142]. Another challenge is to develop a system that summarizes multi-lingual
texts and generates a summary whose quality matches that of a human generated summary
[57]. For multi-document summarizing, redundancy is the biggest problem [15, 239]. The
so far proposed systems strive to identify important sentences in groups of different themes
and hence suffer from the problem of sentence ordering. There is a need for a richer dataset
and computation power. Thus, pre-trained models came into the picture [195]. The hybrid
approach has gained attention recently [221]. By combining extractive and abstractive
techniques, developing efficient hybrid approach methods to generate good quality sum-
maries so that they match closely to human-written ones is another major challenge. The
automatic text summarization evaluation metric such as ROUGE [114] is not considered
complete [147]. The challenge with summary evaluation is to determine how adequate or
useful a summary is relative to its source. Thus, methods for generating and evaluating
summaries should complement each other.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43130 Multimedia Tools and Applications (2023) 82:43089–43144

9.2 Question‑answering

Automatic question–answer generation is a significant advancement still many challenging

issues are yet to be resolved. One such issue is to precisely understand the natural language
questions and deduce the exact meaning to retrieve specific responses [50]. Another chal-
lenge is the selection of a question that has good coverage of the content and is of appropri-
ate difficulty, and in the case of multiple-choice questions, distractor generation is the big
challenge. As input texts grow longer, sequence-to-sequence models struggle to effectively
utilize relevant contexts while avoiding unnecessary information [47]. The models do not
pay much attention to the answers that are critical to question word generation. As a result,
the generated question words do not match the answer form [186]. Previous neural ques-
tion generation models suffer from a problem where a large percentage of the generated
questions contain words from the question target. As a result, they generate unintended
questions [88]. The models are not aware of the positions of the context words. Instead of
considering the close and relevant words to the answer, they copy the context words that
are far apart and irrelevant to the answer [186]. The one most frequent problem is the Lexi-
cal gap between questions. It concerns variation in the formation of questions in natural
language. Users formulate the question in different ways and ask for the same information.
This results in questions that differ lexically but are semantically equivalent. The problem
of word sense ambiguity is still a challenge in the QA field [180]. To leverage the wide
range of the available datasets for the question–answer generation is not trivial. The task
of selecting an appropriate dataset is still an open problem [50]. Challenges also arise due
to the limited size of the user’s utterance, ambiguous, and missing information while inter-
preting a question.

9.3 Dialog response generation

Usually, conversational systems rely on RNN models, and RNNs are not able to model
high-level variability [178]. The end-to-end conversational agents are prone to generat-
ing dull, generic, and boring responses [173]. To elicit a coherent, novel, and insightful
response that is in line with the conversation spectrum, The conversational agents require
adequate, accessible data [178, 181]. However, the conversational models, even with the
powerful performance of neural networks, lack style, which possesses to be an issue as
users may not be entirely satisfied with the interaction. Another problem is to encode con-
textual data such as world facts from knowledge bases or prior conversations. The response
generated has to be contextually relevant to the conversation and also convey accurate
paralinguistic functionality. Generating personalized dialogues is another challenging task
[141].

9.4 Machine translation

Although neural machine translation (NMT) has been witnessing fast-growing research
progresses, there are still many challenges. The major neural MT challenges are listed here.
A major limitation of NMT is that it is not able to incorporable larger contextual infor-
mation efficiently due to the learning ability of the model itself. The problem of reorder-
ing has not been addressed much so far [124]. The problems of alignment mechanism and
vocabulary coverage always affect most of the NMT models [192]. NMT also struggles
to deal with the translation of idioms [3]. Low-resource language MT is another hot spot,

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43131

owing to multiple reasons, including morphological complexity and diversity, in addition

to a lack of resources for many languages [79].

9.5 Story generation

Automatic story generation is challenging since it requires the generation of long-range

dependencies and coherent natural language to describe a sensible sequence of events
[227]. Another challenge in story generation is to create interactive narration along a cer-
tain story path so that the interactor has been provided the ability to modify the space or
even the plot [208]. The commonly observed issues in generated stories are repeated plots,
conflicting logic, and inter-sentence incoherence [20, 34, 53]. Another challenge is to use
constraints to generate a creative story within the structure of the plot [151]. In most sys-
tems, evaluating the topicality, fluency, and overall quality of the stories generated poses a
unique challenge [53].

9.6 Paraphrase generation

The ability to automatically generate alternative phrases of the same content has been dem-
onstrated to be useful in several NLG areas, such as text summarization and question gen-
eration [155]. Automatically generating diverse and accurate paraphrases continues to be a
difficult challenge due to the complexity of natural language [224]. Evaluation of the para-
phrases generated is the most difficult aspect [105]. Another issue to be addressed is the
generation of multiple diverse paraphrases of high quality to enhance generalization and
robustness [160]. The issue of model holistic properties of sentences such as topic, style,
and other features is still challenging.

9.7 Image captioning/ Video description

The major challenge in describing visual information to text is to learn the intermedi-
ate representation between the natural language domain and the visual domain [127].
Another challenge is the fine-grained natural descriptions of images or videos [42]. For
instance, occlusions of interactive objects and unclear unit boundaries present addi-
tional challenges in effectively decoding the intent of the human behavior in a video.
There are challenges associated with automatically generating textual reports for medi-
cal images and helping medical professionals produce reports more accurately and effi-
ciently. The first is to generate prolonged texts with several sentences or even paragraphs,
and the other is to generate captions with a wide range of heterogeneous forms [229].

Discussion on research question 8

The automated text generation applications have various challenges, including the gen-
eration of human-like text that is fluent, unambiguous, and diverse. This section helped to
answer the RQ 8- “What challenges are faced in automated text generation tasks.” As the
textual data is discrete in nature, it takes time and effort to model long-term dependen-
cies while generating consistent text. Thus, applying neural models to different neural text
generation tasks can depend on various factors, including the type of task, the architecture
of the model, the size and quality of the training data, and the evaluation metrics used to

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43132 Multimedia Tools and Applications (2023) 82:43089–43144

measure the performance of the model. It has also been observed that the automated text
generation is challenging, since it is not always possible to reproduce the reported results.
Since the datasets used for text generation models are not always available publicly, it is
difficult to conduct comparisons among various approaches to text generation. The size of
training data and other key hyperparameters have a substantial impact on the quality of the
generation text. However, there are still many open challenges in text generation that need
to be addressed, including the generation of fluent, coherent, diverse, controllable, and con-
sistent human-like text. Inspired by these challenges, the future aspects of this research
area are presented in the next section.

10 Conclusion and future aspects

This survey captures a comprehensive study and up-to-date systematic review of current
advancements in the field of text generation. Text generation applications has been cat-
egorized mainly into three categories: text-to-text generation (T2T), data-to-text generation
(D2T), and multimodality-to-text generation (M2T), depending on the input (data, text, or
multimodal). A variety of text-to-text generation applications, including text summariza-
tion, question–answer generation, story generation, machine translation, dialogue response
generation, and paraphrase generation, have been discussed and analyzed. The main focus
of this survey is on text-to-text generation applications, and it is beyond the scope of this
survey to include all the recent developments in the various data or multimodality-to-text
applications. This paper also mentions various models for text generation, including tra-
ditional and statistical models, deep learning based models, and pre-trained transformer
architectures, and observed that deep learning approaches and transformer-based architec-
tures have been generally achieving better performance than traditional methods. However,
applying these models to different neural text generation tasks can depend on various fac-
tors, including the type of task, the architecture of the model, the size and quality of the
training data, and the evaluation metrics used to measure the performance of the model.
The quality of the generated text has a huge impact on the decoding strategy and optimiza-
tion technique that one employs. This paper discussed the decoding methods and optimi-
zation techniques important for generating human-like fluent text. A diverse text genera-
tion task-specific standard datasets that are required to train, test, and validate the systems
have also been provided in this article, along with their URLs. This field has made much
progress in recent years. As a result, various text generation application-specific tools are
available in the real world, which is also reviewed in this paper along with their strengths
and limitations. Many sectors have started using automated text to improve user experience
as the recent advancements in the technology is capable of generating human-like texts.
Though, the content generated by the automated tools is fast and cheap but lacks artistry
involved in expressing thoughts. The automatic text generation system’s goal is to gener-
ate text as good as human-written text. The assessment of the generated text is essential to
improve the performance of text generative models. However, human evaluation remains
the gold standard for assessing the quality of automated generated texts, but it is time-
consuming and expensive.
For effective text evaluation, various automatic evaluation metrices have been analyzed
and reviewed in this paper. Nevertheless, many open challenges in text generation need to
be addressed, including the generation of fluent, coherent, diverse, controllable, and con-
sistent human-like text. Inspired by these challenges, the future aspects in this research

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43133

direction include generating long-term fluent and coherent text.The advancements in the
generic models should be done so that they can learn from some low resources and can
handlemultiple languages without large quantities of training data. The generation of
diverse texts conditioned by specific attributesand characteristics is another research direc-
tion. Practical applications for real-time text generation should be developed thatensure
responsible usage of the generated text. The need of the hour is to create a universal evalu-
ation metric that suits all textgeneration tasks and reflects all desired properties of text that
correlate with human judgments.
This survey aims to provide a comprehensive overview of current advancements in
automated text generation and to introducethe topic to researchers by providing pointers
and synthesis to pertinent studies. This paper is believed to serve as a valuablereference for
those concerned with learning and advancing this interesting research area.

Declarations I hereby confirm that this work is original and has not been published elsewhere, nor is it cur-
rently under consideration for publication elsewhere. No potential conflict of interest was reported.

Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest.

Ethical approval The author declares that this article complies with the ethical standard.

References
1. Abrishami M, Rashti MJ, Naderan M (2020) Machine Translation Using Improved Attention-based
Transformer with Hybrid Input. In: 2020 6th International Conference on Web Research (ICWR).
IEEE, pp 52–57
2. Acharya M, Kafle K, Kanan C (2018) TallyQA: Answering complex counting questions. arXiv.
https://​doi.​org/​10.​1609/​aaai.​v33i01.​33018​076
3. Agrawal R, Sharma DM (2017) Building an Effective MT System for English-Hindi Using RNN’s.
Int J Artif Intell Appl 8:45–58. https://​doi.​org/​10.​5121/​ijaia.​2017.​8504
4. Alloatti F, Di Caro L, Sportelli G (2019) Real Life Application of a Question Answering System
Using BERT Language Model. In: Proceedings of the 20th Annual SIGdial Meeting on Discourse
and Dialogue. Association for Computational Linguistics, Stroudsburg, PA, USA, pp 250–253
5. Alomari A, Idris N, Sabri AQM, Alsmadi I (2022) Deep reinforcement and transfer learning for
abstractive text summarization: A review. Comput Speech Lang 71:101276. https://​doi.​org/​10.​1016/j.​
csl.​2021.​101276
6. Alsaleh A, Althabiti S, Alshammari I, et al (2022) LK2022 at Qur’an QA 2022: Simple Transformers
Model for Finding Answers to Questions from Qur’an. In: Proceedings ofthe OSACT 2022 Workshop
@LREC2022. Eur Lang Res Assoc (ELRA), Marseille, pp 120–125
7. Ammanabrolu P, Tien E, Cheung W, et al (2019) Guided Neural Language Generation for Automated
Storytelling. 46–55.https://​doi.​org/​10.​18653/​v1/​w19-​3405
8. Anderson P, Fernando B, Johnson M, Gould S (2016) SPICE: Semantic propositional image caption
evaluation. Lect Notes ComputSci (including Subser Lect Notes Artif Intell Lect Notes Bioinformat-
ics) 9909 LNCS:382–398. https://​doi.​org/​10.​1007/​978-3-​319-​46454-1_​24
9. Anderson P, He X, Buehler C, et al (2018) Bottom-Up and Top-Down Attention for Image Captioning
and Visual Question Answering. Proc IEEE Comput Soc Conf Comput Vis Pattern Recognit 6077–
6086.https://​doi.​org/​10.​1109/​CVPR.​2018.​00636
10. Asghar N, Poupart P, Hoey J, et al (2018) Affective neural response generation. Lect Notes Com-
put Sci (including Subser Lect Notes Artif Intell Lect Notes Bioinformatics) 10772 LNCS:154–166.
https://​doi.​org/​10.​1007/​978-3-​319-​76941-7_​12
11. Bahdanau D, Cho K, Bengio Y (2015) Neural Machine Translation by Jointly Learning to Align and
Translate. 3rd Int Conf Learn Represent ICLR 2015 - Conf Track Proc 1–15

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43134 Multimedia Tools and Applications (2023) 82:43089–43144

12. Bapna A, Chen MX, Firat O, et al (2020) Training deeper neural machine translation models with
transparent attention. Proc 2018 Conf Empir Methods Nat Lang Process EMNLP 2018 3028–3033.
https://​doi.​org/​10.​18653/​v1/​d18-​1338
13. Barrull R, Kalita J (2020) Abstractive and mixed summarization for long-single documents. http://​
arxiv.​org/​abs/​20070​1918 1–9
14. Basu S, Ramachandran GS, Keskar NS, Varshney LR (2021) Mirostat: A Neural Text Decoding
Algorithm that Directly Controls Perplexity. ArXiv 200714966:1–25
15. Baumel T, Eyal M, Elhadad M (2018) Query Focused Abstractive Summarization: Incorporating
Query Relevance, Multi-Document Coverage, and Summary Length Constraints into seq2seq Mod-
els. arXiv:180107704. arXiv:​1801.​07704
16. Bengio Y, Simard P, Frasconi P (1994) Learning long-term dependencies with gradient descent is dif-
ficult. IEEE Trans Neural Networks 5:157–166. https://​doi.​org/​10.​1109/​72.​279181
17. Bott S, Saggion H, Figueroa D (2012) A hybrid system for spanish text simplification. 3rd Work
Speech Lang Process Assist Technol SLPAT 2012 2012 Conf North Am Chapter Assoc Comput Lin-
guist Hum Lang Technol NAACL-HLT 2012 - Proc 75–84
18. Bowman SR, Vilnis L, Vinyals O, et al (2016) Generating sentences from a continuous space.
CoNLL 2016 - 20th SIGNLL Conf Comput Nat Lang Learn Proc 10–21. https://​doi.​org/​10.​18653/​v1/​
k16-​1002
19. Bradbury J, Merity S, Xiong C, Socher R (2016) Quasi-Recurrent Neural Networks. 5th Int Conf
Learn Represent 1–11
20. Brown TB, Mann B, Ryder N, et al (2020) Language Models are Few-Shot Learners. Adv Neural Inf
Process Syst
21. Buck C, Bulian J, Ciaramita M, et al (2018) Ask the Right Questions: Active Question Reformulation
with Reinforcement Learning. In: 6th International Conference on Learning Representations, ICLR
2018. Conference Track Proceedings (2018), pp 1–15
22. Cao Z, Luo C, Li W, Li S (2017) Joint Copying and Restricted Generation for Paraphrase. In: 31st
AAAI Conference on Artificial Intelligence, AAAI 2017. AAAI, pp 3152–3158
23. Cao S, Wang L (2021) Controllable Open-ended Question Generation with A New Question Type
Ontology. arXiv 6424–6439. http://​arxiv.​org/​abs/​2107.​00152
24. Celikyilmaz A, Clark E, Gao J (2020) Evaluation of Text Generation: A Survey. 1–75.http://​arxiv.​org/​
abs/​2006.​14799
25. Chen S, Beeferman D, Rosenfeld R (1998) Evaluation metrics for language models. Proc DARPA
Broadcast News Transcr Underst Work 275– 280
26. Chen J, Xiao G, Han X, Chen H (2021) Controllable and Editable Neural Story Plot Generation via
Control-and-Edit Transformer. IEEE Access 9:96692–96699. https://​doi.​org/​10.​1109/​ACCESS.​2021.​
30942​63
27. Chen Y, Xu L, Liu K, et al (2015) Event extraction via dynamic multi-pooling convolutional neural
networks. ACL-IJCNLP 2015 - 53rd Annu Meet Assoc Comput Linguist 7th Int Jt Conf Nat Lang
Process Asian Fed Nat Lang Process Proc Conf 1:167–176. https://​doi.​org/​10.​3115/​v1/​p15-​1017
28. Cheng J, Lapata M (2016) Neural Summarization by Extracting Sentences and Words. In: Proceed-
ings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long
Papers). Association for Computational Linguistics, Stroudsburg, PA, USA, pp 484–494
29. Cho K, van Merriënboer B, Gulcehre C, et al (2014) Learning Phrase Representations using RNN
Encoder–Decoder for Statistical Machine Translation. Empir Methods Nat Lang Process (EMNLP),
Assoc Comput Linguist 1724–1734. https://​doi.​org/​10.​1128/​jcm.​28.9.​2159-.​1990
30. Cho WS, Zhang Y, Rao S, et al (2021) Contrastive Multi-document Question Generation. In: Pro-
ceedings of the 16th Conference of the European Chapter of the Association for Computational Lin-
guistics: Main Volume. Association for Computational Linguistics, Stroudsburg, PA, USA, pp 12–30
31. Chung J, Gulcehre C, Cho K, Bengio Y (2015) Gated Feedback Recurrent Neural Networks. In: 32nd
International Conference on Machine Learning, ICML 2015. ICML
32. Chung J, Gulcehre C, Cho K, Bengio Y (2014) Empirical Evaluation of Gated Recurrent Neural Net-
works on Sequence Modeling. ArXiv 14123555:1–9
33. Clark E, Celikyilmaz A, Smith NA (2020) Sentence mover’s similarity: Automatic evaluation for
multi-sentence texts. ACL 2019 - 57th Annu Meet Assoc Comput Linguist Proc Conf 2748–2760.
https://​doi.​org/​10.​18653/​v1/​p19-​1264
34. Clark E, Ji Y, Smith NA (2018) Neural Text Generation in Stories Using Entity Representations
as Context. In: Proceedings of the 2018 Conference of the North American Chapter of the Asso-
ciation for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers).
Association for Computational Linguistics, Stroudsburg, PA, USA, pp 2250–2260

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43135

35. Clinchant S, Jung KW, Nikoulina V (2019) On the use of BERT for Neural Machine Translation.
In: Proceedings of the 3rd Workshop on Neural Generation and Translation. Association for Com-
putational Linguistics, Stroudsburg, PA, USA, pp 108–117
36. Cui Q, Wu S, Liu Q et al (2020) MV-RNN: A Multi-View Recurrent Neural Network for Sequen-
tial Recommendation. IEEE Trans Knowl Data Eng 32:317–331. https://​doi.​org/​10.​1109/​TKDE.​
2018.​28812​60
37. Dai B, Fidler S, Urtasun R, Lin D (2017) Towards Diverse and Natural Image Descriptions via
a Conditional GAN. Proc IEEE Int Conf Comput Vis 2017-Octob:2989–2998. https://​doi.​org/​10.​
1109/​ICCV.​2017.​323
38. Dauphin YN, Fan A, Auli M, Grangier D (2017) Language modeling with gated convolutional
networks. 34th Int Conf Mach Learn ICML 2017 2:1551–1559
39. Denil M, Demiraj A, Kalchbrenner N et al (2014) Modelling, Visualising and Summarising Docu-
ments with a Single Convolutional Neural Network. ArXiv 14063830:1–10
40. Devlin J, Chang M, Lee K, Toutanova K (2019) BERT: Pre-training of Deep Bidirectional Trans-
formers for Language Understanding. arXiv:181004805v2 4171–4186. https://​doi.​org/​10.​18653/​
v1/​N19-​1423
41. Dinan E, Roller S, Shuster K, et al (2019) Wizard of Wikipedia: Knowledge-Powered Conversa-
tional agents. In: ICLR. pp 1–18
42. Donahue J, Hendricks LA, Rohrbach M et al (2017) Long-Term Recurrent Convolutional Net-
works for Visual Recognition and Description. IEEE Trans Pattern Anal Mach Intell 39:677–691.
https://​doi.​org/​10.​1109/​TPAMI.​2016.​25991​74
43. Dong L, Mallinson J, Reddy S, Lapata M (2017) Learning to paraphrase for question answering.
arXiv 875–886
44. Dong L, Wei F, Zhou M, Xu K (2015) Question Answering over Freebase with Multi-Column
Convolutional Neural Networks. In: Proceedings of the 53rd Annual Meeting of the Association
for Computational Linguistics and the 7th International Joint Conference on Natural Language
Processing (Volume 1: Long Papers). Association for Computational Linguistics, Stroudsburg,
PA, USA, pp 260–269
45. Dozat T (2016) INCORPORATING NESTEROV MOMENTUM INTO ADAM. In: ICLR Work-
shop. ICLR, pp 2013–2016
46. Du X, Cardie C (2017) Identifying Where to Focus in Reading Comprehension for Neural Question
Generation. In: Proceedings of the 2017 Conference on Empirical Methods in Natural Language Pro-
cessing. Association for Computational Linguistics, Stroudsburg, PA, USA, pp 2067–2073
47. Du X, Shao J, Cardie C (2017) Learning to Ask: Neural Question Generation for Reading Com-
prehension. arXiv:170500106v1
48. Duan N, Tang D, Chen P, Zhou M (2017) Question Generation for Question Answering. In: Pro-
ceedings of the 2017 Conference on Empirical Methods in Natural Language Processing. Associa-
tion for Computational Linguistics, Stroudsburg, PA, USA, pp 866–874
49. Duchi JC, Bartlett PL, Wainwright MJ (2012) Randomized smoothing for (parallel) stochastic optimi-
zation. In: 2012 IEEE 51st IEEE Conference on Decision and Control (CDC). IEEE, pp 5442–5444
50. Dwivedi SK, Singh V (2013) Research and Reviews in Question Answering System. Procedia
Technol 10:417–424. https://​doi.​org/​10.​1016/j.​protcy.​2013.​12.​378
51. Evans R, Grefenstette E (2018) Learning Explanatory Rules from Noisy Data. J Artif Intell Res
61:1–64. https://​doi.​org/​10.​1613/​jair.​5714
52. Faizan A, Lohmann S (2018) Automatic generation of multiple choice questions from slide con-
tent using linked data. ACM Int Conf Proceeding Ser doi 10(1145/3227609):3227656
53. Fan A, Lewis M, Dauphin Y (2018) Hierarchical neural story generation. ACL 2018 - 56th Annu
Meet AssocComput Linguist Proc Conf (Long Pap 1:889–898. https://​doi.​org/​10.​18653/​v1/​p18-​1082
54. Feng B, Liu D, Sun Y (2021) Evolving transformer architecture for neural machine translation. In:
Proceedings of the Genetic and Evolutionary Computation Conference Companion. ACM, New
York, NY, USA, pp 273–274
55. Frome A, Corrado GS, Shelens J et al (2018) DeViSE: A Deep Visual-Semantic Embedding
Model Andrea. Phys C Supercond its Appl. https://​doi.​org/​10.​1016/​0921-​4534(95)​00110-7
56. Fung P, Bertero D, Xu P, et al (2014) Empathetic Dialog Systems. In: The International Confer-
ence on Language Resources and Evaluation. European Language Resources Association. Euro-
pean Language Resources Association
57. Gambhir M, Gupta V (2017) Recent automatic text summarization techniques. Artif Intell Rev 47:1–
66. https://​doi.​org/​10.​1007/​s10462-​016-​9475-9

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43136 Multimedia Tools and Applications (2023) 82:43089–43144

58. Gao P, Li H, Li S, et al (2018) Question-Guided Hybrid Convolution for Visual Question Answering.
Lect Notes Comput Sci (including Subser Lect Notes Artif Intell Lect Notes Bioinformatics) 11205
LNCS:485–501. https://​doi.​org/​10.​1007/​978-3-​030-​01246-5_​29
59. Garbacea C, Mei Q (2020) Neural Language Generation: Formulation, Methods, and Evaluation.
http://​arxiv.​org/​abs/​20071​5780
60. Gardent C, Kow E (2007) A symbolic approach to near-deterministic surface realisation using tree
adjoining grammar. In: ACL 2007 - Proceedings of the 45th Annual Meeting of the Association for
Computational Linguistics. pp 328–335
61. Gehring J, Auli M, Grangier D, et al (2017) Convolutional sequence to sequence learning. 34th Int
Conf Mach Learn ICML 2017 3:2029–2042
62. Goldberg Y (2016) A Primer on Neural Network Models for Natural Language Processing. J Artif
Intell Res 57:345–420. https://​doi.​org/​10.​1613/​jair.​4992
63. Goyal T, Li JJ, Durrett G (2022) News Summarization and Evaluation in the Era of GPT-3. arXiv.
http://​arxiv.​org/​abs/​2209.​12356
64. Grechishnikova D (2021) Transformer neural network for protein-specific de novo drug generation as
a machine translation problem. Sci Rep 11:321. https://​doi.​org/​10.​1038/​s41598-​020-​79682-4
65. Gu J, Bradbury J, Xiong C, et al (2017) Non-Autoregressive Neural Machine Translation. Proc 2018
Conf Empir Methods Nat Lang Process 479–488
66. Guan J, Wang Y, Huang M (2019) Story Ending Generation with Incremental Encoding and Com-
monsense Knowledge. Proc AAAI Conf Artif Intell 33:6473–6480. https://​doi.​org/​10.​1609/​aaai.​
v33i01.​33016​473
67. Gupta A, Agarwal A, Singh P, Rai P (2018) A deep generative framework for paraphrase generation.
32nd AAAI Conf Artif Intell AAAI 2018 5149–5156
68. Harrison B, Purdy C, Riedl MO (2021) Toward Automated Story Generation with Markov Chain
Monte Carlo Methods and Deep Neural Networks. In: AAAI Conference on Artificial Intelligence
and Interactive Digital Entertainment. AAAI, pp 191–197
69. Harrison V, Walker M (2018) Neural generation of diverse questions using answer focus, contextual
and linguistic features. In: Proceedings ofThe 11th International Natural Language Generation Con-
ference. Association for Computational Linguistics, Tilburg, The Netherlands, pp 296–306
70. Hashimoto TB, Zhang H, Liang P (2019) Unifying human and statistical evaluation for natural lan-
guage generation. NAACL HLT 2019 - 2019 Conf North Am Chapter Assoc Comput Linguist Hum
Lang Technol - Proc Conf 1:1689–1701. https://​doi.​org/​10.​18653/​v1/​n19-​1169
71. He X, Deng L (2017) Deep Learning for VisuaLunDerstanDing Deep Learning for Image-to-Text
Generation. IEEE Signal Process Mag 109–116. https://​doi.​org/​10.​1109/​MSP.​2017.​27415​10
72. Helcl J, Haddow B, Birch A (2022) Non-Autoregressive Machine Translation: It’s Not as Fast as it
Seems. In: Proceedings of the 2022 Conference of the North American Chapter of the Association for
Computational Linguistics: Human Language Technologies. Association for Computational Linguis-
tics, Stroudsburg, PA, USA, pp 1780–1790
73. Hidasi B, Quadrana M, Karatzoglou A, Tikk D (2016) Parallel Recurrent Neural Network Architec-
tures for Feature-rich Session-based Recommendations. In: Proceedings of the 10th ACM Conference
on Recommender Systems. ACM, New York, NY, USA, pp 241–248
74. Hochreiter S, Schmidhuber J (21997) Long Short-Term Memory. Neural Comput. https://​doi.​org/​10.​
17582/​journ​al.​pjz/​2018.​50.6.​2199.​2207
75. Holtzman A, Buys J, Du L, et al (2019) The Curious Case of Neural Text Degeneration. CEUR Work-
shop Proc 2540:
76. Huang C, Zaïane OR, Trabelsi A, Dziri N (2018) Automatic dialogue generation with expressed emo-
tions. NAACL HLT 2018 - 2018 Conf North Am Chapter AssocComput Linguist Hum Lang Technol
- Proc Conf 2:49–54. https://​doi.​org/​10.​18653/​v1/​n18-​2008
77. Iyyer M, Wieting J, Gimpel K, Zettlemoyer L (2018) Adversarial example generation with syntac-
tically controlled paraphrase networks. NAACL HLT 2018 - 2018 Conf North Am Chapter Assoc
Comput Linguist Hum Lang Technol - Proc Conf 1:1875–1885. https://​doi.​org/​10.​18653/​v1/​n18-​1170
78. Jain P, Agrawal P, Mishra A, et al (2017) Story Generation from Sequence of Independent Short
Descriptions. ArXiv.https://​doi.​org/​10.​48550/​arXiv.​1707.​05501
79. Jha S, Sudhakar A, Singh AK (2018) Learning cross-lingual phonological and orthagraphic adapta-
tions: A case study in improving neural machine translation between low-resource languages. arXiv
1–48. https://​doi.​org/​10.​15398/​jlm.​v7i2.​214
80. Jin J, Fu K, Cui R, et al (2015) Aligning where to see and what to tell: image caption with region-
based attention and scene factorization. 1–20
81. Jozefowicz R, Vinyals O, Schuster M, et al (2016) Exploring the Limits of Language Modeling.
arXiv:160202410

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43137

82. Kalchbrenner N, Blunsom P (2013) Recurrent continuous translation models. EMNLP 2013 -
2013 Conf Empir Methods Nat Lang Process Proc Conf 1700–1709
83. Kamal Eddine M, Shang G, Tixier A, Vazirgiannis M (2022) FrugalScore: Learning Cheaper,
Lighter and Faster Evaluation Metrics for Automatic Text Generation. In: Proceedings of the 60th
Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Asso-
ciation for Computational Linguistics, Stroudsburg, PA, USA, pp 1305–1318
84. Kannan A, Kurach K, Ravi S, et al (2016) Smart Reply. In: Proceedings of the 22nd ACM SIG-
KDD International Conference on Knowledge Discovery and Data Mining. ACM, New York, NY,
USA, pp 955–964
85. Karpathy A, Joulin A, Fei-Fei L (2014) Deep Fragment Embeddings for Bidirectional Image
Sentence Mapping. 27th International Conference on Neural Information Processing Systems
(NIPS’14). MIT Press, Cambridge, MA, USA, pp 1889–1897
86. Keneshloo Y, Shi T, Ramakrishnan N et al (2020) Deep Reinforcement Learning for Sequence-to-
Sequence Models 31:2469–2489
87. Khamparia A, Pandey B, Tiwari S et al (2020) An Integrated Hybrid CNN–RNN Model for Visual
Description and Generation of Captions. Circuits, Syst Signal Process 39:776–788. https://​doi.​org/​
10.​1007/​s00034-​019-​01306-8
88. Kim Y, Lee H, Shin J, Jung K (2019) Improving Neural Question Generation Using Answer Sepa-
ration. Thirty-Third AAAI Conf Artif Intell Improv
89. Kingma DP, Ba J (2015) Adam: A Method for Stochastic Optimization. In: 3rd International Con-
ference on Learning Representations. ICLR 2015, pp 1–15
90. Kiros R, Salakhutdinov R, Zemel R (2014) Multimodal neural language models. 31st Int Conf
Mach Learn ICML 2014 3:2012–2025
91. Kiros R, Salakhutdinov R, Zemel RS (2014) Unifying Visual-Semantic Embeddings with Multi-
modal Neural Language Models. ArXiv 14112539:1–13
92. Kitchenham B, Pearl Brereton O, Budgen D et al (2009) Systematic literature reviews in software
engineering - A systematic literature review. Inf Softw Technol 51:7–15. https://​doi.​org/​10.​1016/j.​
infsof.​2008.​09.​009
93. Knight K, Marcu D (2000) Statistics-Based Summarization - Step One: Sentence Compression.
In: Knight2000StatisticsBasedS. American Association for Artificial Intelligence (www.​aaai.​org),
pp 703–710
94. Kumar A, Irsoy O, Ondruska P, et al (2016) Ask me anything: Dynamic memory networks for
natural language processing. 33rd Int Conf Mach Learn ICML 2016 3:2068–2078
95. Kumar V, Ramakrishnan G, Li YF (2019) Putting the horse before the cart: A generator-evaluator
framework for question generation from text. CoNLL 2019 - 23rd ConfComput Nat Lang Learn
Proc Conf 812–821. https://​doi.​org/​10.​18653/​v1/​k19-​1076
96. Lavie A, Agarwal A (2005) METEOR: An automatic metric for MT evaluation with high lev-
els of correlation with human judgments. In: Proceedings of the Second Workshop on Statistical
Machine Translation
97. LeCun Y, Bengio Y, Hinton G (2015) Deep learning. Nature 521:436–444. https://​doi.​org/​10.​
1038/​natur​e14539
98. Lee J, Liang B, Fong H (2021) Restatement and Question Generation for Counsellor Chatbot. In:
Proceedings of the 1st Workshop on NLP for Positive Impact. Association for Computational Lin-
guistics, Stroudsburg, PA, USA, pp 1–7
99. Lee J, Yoon W, Kim S, et al (2019) BioBERT: a pre-trained biomedical language representation
model for biomedical text mining. Bioinformatics 1–7. https://​doi.​org/​10.​1093/​bioin​forma​tics/​
btz682
100. Lelkes AD, Tran VQ, Yu C (2021) Quiz-Style Question Generation for News Stories. In: Proceed-
ings of the Web Conference 2021. ACM, New York, NY, USA, pp 2501–2511
101. Lemberger P (2020) Deep Learning Models for Automatic Summarization. http://​arxiv.​org/​abs/​
20051​1988 1–13
102. Lewis M, Liu Y, Goyal N, et al (2020) BART: Denoising Sequence-to-Sequence Pre-training for
Natural Language Generation, Translation, and Comprehension. 7871–7880. https://​doi.​org/​10.​
18653/​v1/​2020.​acl-​main.​703
103. Lewis M, Yarats D, Dauphin YN, et al (2017) Deal or no deal? End-to-end learning for negotiation
dialogues.EMNLP 2017 - Conf Empir Methods Nat Lang Process Proc 2443–2453. https://​doi.​
org/​10.​18653/​v1/​d17-​1259
104. Li J, Galley M, Brockett C, et al (2016) A diversity-promoting objective function for neural conversa-
tion models. 2016 Conf North Am Chapter Assoc Comput Linguist Hum Lang Technol NAACL HLT
2016 - Proc Conf 110–119. https://​doi.​org/​10.​18653/​v1/​n16-​1014

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43138 Multimedia Tools and Applications (2023) 82:43089–43144

105. Li Z, Jiang X, Shang L, Li H (2018) Paraphrase Generation with Deep Reinforcement Learning. In:
Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. Associa-
tion for Computational Linguistics, Stroudsburg, PA, USA, pp 3865–3878
106. Li B, Lee-urban S, Johnston G, Riedl MO (2013) Story Generation with Crowdsourced Plot Graphs.
AAAI, pp 598–604
107. Li Y, Li K, Ning H, et al (2021) Towards an Online Empathetic Chatbot with Emotion Causes. In:
Proceedings of the 44th International ACM SIGIR Conference on Research and Development in
Information Retrieval. ACM, New York, NY, USA, pp 2041–2045
108. Li J, Luong MT, Jurafsky D (2015) A hierarchical neural Autoencoder for paragraphs and documents.
ACL-IJCNLP 2015 - 53rd Annu Meet Assoc Comput Linguist 7th Int Jt Conf Nat Lang Process
Asian Fed Nat Lang Process Proc Conf 1:1106–1115. https://​doi.​org/​10.​3115/​v1/​p15-​1107
109. Li J, Monroe W, Jurafsky D (2016) A Simple, Fast Diverse Decoding Algorithm for Neural Genera-
tion. ArXiv: 161108562
110. Li J, Monroe W, Ritter A, et al (2016) Deep reinforcement learning for dialogue generation. EMNLP
2016 - Conf Empir Methods Nat Lang Process Proc 1192–1202. https://​doi.​org/​10.​18653/​v1/​
d16-​1127
111. Li J, Monroe W, Shi T, et al (2017) Adversarial learning for neural dialogue generation. EMNLP
2017 - Conf Empir Methods Nat Lang Process Proc 2157–2169.https://​doi.​org/​10.​18653/​v1/​d17-​1230
112. Li S, Tao Z, Li K, Fu Y (2019) Visual to Text: Survey of Image and Video Captioning. IEEE Trans
Emerg Top Comput Intell 3:297–312. https://​doi.​org/​10.​1109/​TETCI.​2019.​28927​55
113. Liao K, Lebanoff L, Liu F (2018) Abstract Meaning Representation for Multi-Document Summariza-
tion. In: International Conference on Computational Linguistics. Santa Fe, New Mexico, USA, pp
1178–1190
114. Lin C-Y (2004) ROUGE: A Package for Automatic Evaluation of Summaries. In: Text Summariza-
tion Branches Out. Association for Computational Linguistics, Barcelona, Spain, pp 74–81
115. Liu P, Huang C, Mou L (2022) Learning Non-Autoregressive Models from Search for Unsupervised
Sentence Summarization. arXiv 7916–7929. https://​doi.​org/​10.​18653/​v1/​2022.​acl-​long.​545
116. Liu Y, Lapata M (2019) Text Summarization with Pretrained Encoders. arXiv. http://​arxiv.​org/​abs/​
1908.​08345
117. Liu X, Lei W, Lv J, Zhou J (2022) Abstract Rule Learning for Paraphrase Generation. In: Proceedings
of the Thirty-First International Joint Conference on Artificial Intelligence. International Joint Con-
ferences on Artificial Intelligence Organization, California, pp 4273–4279
118. Liu PJ, Saleh M, Pot E, et al (2018) Generating Wikipedia by Summarizing Long Sequences. In: 6th
International Conference on Learning Representations, ICLR 2018 - Conference Track Proceedings.
pp 1–18
119. Liu J, Shen D, Zhang Y, et al (2021) What Makes Good In-Context Examples for GPT-$3$? DeeLIO
2022 - Deep Learn Insid Out 3rd Work Knowl Extr Integr Deep Learn Archit Proc Work 3:100–114.
http://​arxiv.​org/​abs/​2101.​06804
120. Liu W, Wang Z, Liu X et al (2017) A survey of deep neural network architectures and their applica-
tions. Neurocomputing 234:11–26. https://​doi.​org/​10.​1016/j.​neucom.​2016.​12.​038
121. Lopyrev K (2015) Generating News Headlines with Recurrent Neural Networks. ArXiv
151201712:1–9
122. Lu J, Yang J, Batra D, Parikh D (2016) Hierarchical Question-Image Co-Attention for Visual Ques-
tion Answering. Adv Neural Inf Process Syst 289–297
123. Lu S, Zhu Y, Zhang W, et al (2018) Neural Text Generation: Past, Present and Beyond. http://​arxiv.​
org/​abs/​18030​7133
124. Luong M-T, Pham H, D. Manning C (2015) Effective Approaches to Attention-based Neural Machine
Translation. In: Proceedings ofthe 2015 Conference on Empirical Methods in Natural Language Pro-
cessing. Lisbon, pp 1412–1421
125. Ma S, Sun X, Li W, et al (2018) Query and output: Generating words by querying distributed word
representations for paraphrase generation. NAACL HLT 2018 - 2018 Conf North Am Chapter Assoc
Comput Linguist Hum Lang Technol - Proc Conf 1:196–206. https://​doi.​org/​10.​18653/​v1/​n18-​1018
126. Makav B, Kilic V (2019) A New Image Captioning Approach for Visually Impaired People. ELECO
2019 - 11th Int Conf Electr Electron Eng 945–949. https://​doi.​org/​10.​23919/​ELECO​47770.​2019.​
89906​30
127. Mao J, Xu W, Yang Y, et al (2015) Deep captioning with multimodal recurrent neural networks
(m-RNN). 3rd Int Conf Learn Represent ICLR 2015 - Conf Track Proc 1090:1–17
128. Martin LJ, Ammanabrolu P, Wang X, et al (2018) Event representations for automated story genera-
tion with deep neural nets. 32nd AAAI Conf Artif Intell AAAI 2018 868–875

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43139

129. Mehta P, Arora G, Majumder P (2018) Attention based Sentence Extraction from Scientific Articles
using Pseudo-Labeled data. Assoc Comput Mach 2–5. https://​doi.​org/​10.​48550/​arXiv.​1802.​04675
130. Michalopoulos G, Chen H, Wong A (2020) Where’s the Question? A Multi-channel Deep Convolu-
tional Neural Network for Question Identification in Textual Data.215–226. https://​doi.​org/​10.​18653/​
v1/​2020.​clini​calnlp-​1.​24
131. Mou L, Song Y, Yan R, et al (2016) Sequence to Backward and Forward Sequences: A Content-
Introducing Approach to Generative Short-Text Conversation. In: COLING 2016 - 26th International
Conference on Computational Linguistics, Proceedings of COLING 2016: Technical Papers. COL-
ING, pp 3349–3358
132. Mridha MF, Lima AA, Nur K et al (2021) A Survey of Automatic Text Summarization: Progress,
Process and Challenges. IEEE Access 9:156043–156070. https://​doi.​org/​10.​1109/​ACCESS.​2021.​
31297​86
133. Nag D, Das B, Dash PS, et al (2015) From word embeddings to document distances. In: 32nd Inter-
national Conference on International Conference on Machine Learning. ICML’15, Lille, France, pp
957–966
134. Nallapati R, Zhai F, Zhou B (2017) SummaRuNNer: A Recurrent Neural Network based Sequence
Model for Extractive Summarization of Documents. In: 31st AAAI Conference on Artificial Intel-
ligence, AAAI 2017. pp 3075–3081
135. Nallapati R, Zhou B, dos Santos C, et al (2016) Abstractive Text Summarization using Sequence-to-
sequence RNNs and Beyond. In: Proceedings of The 20th SIGNLL Conference on Computational
Natural Language Learning. Association for Computational Linguistics, Stroudsburg, PA, USA, pp
280–290
136. Narayan S, Cohen SB, Lapata M (2018) Ranking Sentences for Extractive Summarization with Rein-
forcement Learning. In: Proceedings of the 2018 Conference of the North American Chapter of the
Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers).
Association for Computational Linguistics, Stroudsburg, PA, USA, pp 1747–1759
137. Narayan S, Gardent C (2012) Structure-Driven Lexicalist Generation. 24th International Confer- ence
in Computational Linguistics (COLING). Mumbai, India, pp 100–113
138. Narayan S, Gardent C (2020) Deep Learning Approaches to Text Production. Synth Lect Hum Lang
Technol 13:1–199. https://​doi.​org/​10.​2200/​S0097​9ED1V​01Y20​1912H​LT044
139. Narayan S, Gardent C, Narayan S, et al (2015) Hybrid Simplification using Deep Semantics and
Machine Translation To cite this version : HAL Id : hal-01109581
140. Nguyen A (2021) Language Model Evaluation in Open-ended Text Generation. arXiv. http://​arxiv.​
org/​abs/​2108.​03578
141. Niu T, Bansal M (2018) Polite dialogue generation without parallel data. arXiv. https://​doi.​org/​10.​
1162/​tacl_a_​00027
142. PadmaPriya G, Duraiswamy K (2014) AN APPROACH FOR TEXT SUMMARIZATION USING
DEEP LEARNING ALGORITHM. J Comput Sci 10:1–9. https://​doi.​org/​10.​3844/​jcssp.​2014.1.9
143. Papineni K, Roukos S, Ward T, Zhu W-J (2002) BLEU: a Method for Automatic Evaluation of
Machine Translation. In: 40th Annual Meeting of the Association for Computational Linguistics
(ACL). ACL, pp 311–318
144. Park HJ, Lee JS, Ko JG (2020) Achieving Real-Time Sign Language Translation Using a Smart-
phone’s True Depth Images. In: 12th International Conference on Communication Systems & Net-
works (COMSNETS). IEEE, pp 622–625
145. Parveen D, Mesgar M, Strube M (2016) Generating coherent summaries of scientific articles using
coherence patterns. EMNLP 2016 - Conf Empir Methods Nat Lang Process Proc 772–783. https://​
doi.​org/​10.​18653/​v1/​d16-​1074
146. Pascanu R, Mikolov T, Bengio Y (2018) On the difficulty of training recurrent neural networks. Phy-
logenetic Divers Appl Challenges Biodivers Sci 41–71. https://​doi.​org/​10.​1007/​978-3-​319-​93145-6_3
147. Paulus R, Xiong C, Socher R (2017) A Deep Reinforced Model for Abstractive Summarization. 6th
Int Conf Learn Represent ICLR 2018 - Conf Track Proc 1–12
148. Pauws S, Gatt A, Krahmer E, Reiter E (2019) Making effective use of healthcare data using
data-to-text technology. Data Sci Healthc Methodol Appl 119–145. https://​doi.​org/​10.​1007/​
978-3-​030-​05249-2_4
149. Pawade D, Sakhapara A, Jain M et al (2018) Story Scrambler - Automatic Text Generation Using
Word Level RNN-LSTM. Int J Inf Technol Comput Sci 10:44–53. https://​doi.​org/​10.​5815/​ijitcs.​2018.​
06.​05
150. Pedersoli M, Lucas T, Schmid C, Verbeek J (2017) Areas of Attention for Image Captioning. Proc
IEEE Int Conf Comput Vis 2017-Octob:1251–1259. https://​doi.​org/​10.​1109/​ICCV.​2017.​140

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43140 Multimedia Tools and Applications (2023) 82:43089–43144

151. Peng N, Ghazvininejad M, May J, Knight K (2018) Towards Controllable Story Generation. In:
Proceedings of the First Workshop on Storytelling. Association for Computational Linguistics,
Stroudsburg, PA, USA, pp 43–49
152. Peng B, Yao K (2015) Recurrent Neural Networks with External Memory for Language Under-
standing. arXiv:150600195v1
153. Peng D, Zhou M, Liu C, Ai J (2020) Human–machine dialogue modelling with the fusion of word-
and sentence-level emotions. Knowledge-Based Syst 192:105319. https://​doi.​org/​10.​1016/j.​kno-
sys.​2019.​105319
154. Portet F, Reiter E, Gatt A et al (2009) Automatic generation of textual summaries from neonatal
intensive care data. Artif Intell 173:789–816. https://​doi.​org/​10.​1016/j.​artint.​2008.​12.​002
155. Prakash A, Hasan SA, Lee K, et al (2016) Neural paraphrase generation with stacked residual
LSTM Networks. COLING 2016 - 26th Int Conf Comput Linguist Proc COLING 2016 Tech Pap
2923–2934
156. Przybocki M, Peterson K, Bronsart S, Sanders G (2009) The NIST 2008 metrics for machine trans-
lation challenge-overview, methodology, metrics, and results. Mach Transl 23:71–103. https://​doi.​
org/​10.​1007/​s10590-​009-​9065-6
157. Qi W, Gong Y, Jiao J, et al (2021) BANG: Bridging Autoregressive and Non-autoregressive Gen-
eration with Large Scale Pretraining
158. Qian N (1999) On the momentum term in gradient descent learning algorithms. Neural Netw
12:145–151. https://​doi.​org/​10.​1016/​S0893-​6080(98)​00116-6
159. Qian Q, Huang M, Zhao H, et al (2018) Assigning personality/identity to a chatting machine for
coherent conversation generation. Proc Twenty-Seventh Int Jt Conf Artif Intell 4279–4285
160. Qian L, Qiu L, Zhang W, et al (2019) Exploring Diverse Expressions for Paraphrase Generation.
In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing
and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP).
Association for Computational Linguistics, Stroudsburg, PA, USA, pp 3171–3180
161. Qian L, Zhou H, Bao Y, et al (2021) Glancing Transformer for Non-Autoregressive Neural
Machine Translation. In: Proceedings of the 59th Annual Meeting of the Association for Compu-
tational Linguistics and the 11th International Joint Conference on Natural Language Processing
(Volume 1: Long Papers). Association for Computational Linguistics, Stroudsburg, PA, USA, pp
1993–2003
162. Radford A, Narasimhan K (2018) Improving Language Understanding by Generative Pre-Training
163. Raffel C, Shazeer N, Roberts A, et al (2020) Exploring the Limits of Transfer Learning with a
Unified Text-to-Text Transformer. arXiv:191010683v2 1–67
164. Rajasekar AA, Garera N (2021) Answer Generation for Questions With Multiple Information
Sources in E-Commerce. Proc Flip DS Conf 1:
165. Rajpurkar P, Zhang J, Lopyrev K, Liang P (2016) SQuAD: 100,000+ Questions for Machine
Comprehension of Text. Proc 2016 Conf Empir Methods Nat Lang Process 2383–2392.https://​doi.​
org/​10.​18653/​v1/​D16-​1264
166. Ranzato M, Chopra S, Auli M, Zaremba W (2016) Sequence Level Training with Recurrent Neural
Networks. In: 4th International Conference on Learning Representations, ICLR. ICLR, pp 1–16
167. Rashkin H, Smith EM, Li M, Boureau YL (2020) Towards empathetic open-domain conversation
models: A new benchmark and dataset. ACL 2019 - 57th Annu Meet Assoc Comput Linguist Proc
Conf 5370–5381. https://​doi.​org/​10.​18653/​v1/​p19-​1534
168. Reiter E, Dale R (1997) Building applied natural language generation systems. Nat Lang Eng
3:57–87. https://​doi.​org/​10.​1017/​S1351​32499​70015​02
169. Ren Z, Wang X, Zhang N, et al (2017) Deep reinforcement learning-based image captioning with
embedding reward. Proc - 30th IEEE Conf Comput Vis Pattern Recognition, CVPR 2017 2017-
Janua:1151–1159. https://​doi.​org/​10.​1109/​CVPR.​2017.​128
170. Roemmele M (2016) Writing Stories with Help from Recurrent Neural Networks. In: Proceedings of
the Thirtieth AAAI Conference on Artificial Intelligence (AAAI-16) Writing. AAAI, pp 4311–4312
171. Roemmele M, Gordon AS (2015) Interactive Storytelling. Springer International Publishing, Cham
172. Rush AM, Chopra S, Weston J (2015) A Neural Attention Model for Abstractive Sentence
Summarization. In: Proceedings of the 2015 Conference on Empirical Methods in Natural
Language Processing. Association for Computational Linguistics, Stroudsburg, PA, USA, pp
379–389
173. Santhanam S, Shaikh S (2019) A Survey of Natural Language Generation Techniques with a
Focus on Dialogue Systems - Past. Present and Future Directions A Survey of Natural Language
Generation Techniques with a Focus on Dialogue Systems - Past, Present and Future Directions.
https://​doi.​org/​10.​5087/​dad.​DOINU​MBER

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43141

174. Saxena SS, Saranya G, Aggarwal D (2020) A Convolutional Recurrent Neural Network ( CRNN
) Based Approach for Text Recognition and Conversion of Text To Speech in Various Indian Lan-
guages. Int J Adv Sci Technol 29:2770–2776
175. Schuster M, Paliwal KK (1997) Bidirectional recurrent neural networks. IEEE Trans Signal Process
45:2673–2681. https://​doi.​org/​10.​1109/​78.​650093
176. Scialom T, Hill F (2021) BEAMetrics: A Benchmark for Language Generation Evaluation Evalua-
tion. arXiv 1–20
177. See A, Liu PJ, Manning CD (2017) Get To The Point: Summarization with Pointer-Generator Net-
works. In: Proceedings of the 55th Annual Meeting of the Association for Computational Linguis-
tics (Volume 1: Long Papers). Association for Computational Linguistics, Stroudsburg, PA, USA, pp
1073–1083
178. Serban I V., Sordoni A, Bengio Y, et al (2016) Building End-To-End Dialogue Systems Using Gen-
erative Hierarchical Neural Network Models. In: 30th AAAI Conference on Artificial Intelligence,
AAAI 2016. AAAI Press, pp 3776–3783
179. Shetty R, Rohrbach M, Hendricks LA, et al (2017) Speaking the Same Language: Matching Machine
to Human Captions by Adversarial Training. Proc IEEE Int Conf Comput Vis 2017-Octob:4155–
4164. https://​doi.​org/​10.​1109/​ICCV.​2017.​445
180. Song L, Wang Z, Hamza W, et al (2018) Leveraging Context Information for Natural Question Gen-
eration. In: Proceedings of the 2018 Conference of the North American Chapter of the Association
for Computational Linguistics: Human Language Technologies, Volume 2 (Short Papers). Associa-
tion for Computational Linguistics, Stroudsburg, PA, USA, pp 569–574
181. Sordoni A, Galley M, Auli M, et al (2015) A Neural Network Approach to Context-Sensitive Gen-
eration of Conversational Responses. In: Proceedings of the 2015 Conference of the North American
Chapter of the Association for Computational Linguistics: Human Language Technologies. Associa-
tion for Computational Linguistics, Stroudsburg, PA, USA, pp 196–205
182. Sriram A, Jun H, Satheesh S, Coates A (2018) Cold fusion: Training Seq2seq models together with
language models. Proc AnnuConf Int Speech Commun Assoc INTERSPEECH 2018-Septe:387–391.
https://​doi.​org/​10.​21437/​Inter​speech.​2018-​1392
183. Stasaski K, Rathod M, Tu T, et al (2021) Automatically Generating Cause-and-Effect Questions from
Passages. Proc 16th Work Innov Use NLP Build Educ Appl BEA 2021 - held conjunction with 16th
Conf Eur Chapter Assoc Comput Linguist EACL 2021 158–170
184. Su Y, Wang Y, Cai D et al (2021) PROTOTYPE-TO-STYLE: Dialogue Generation with Style-Aware
Editing on Retrieval Memory. IEEE/ACM Trans Audio Speech Lang Process 29:2152–2161. https://​
doi.​org/​10.​1109/​TASLP.​2021.​30879​48
185. Subramanian S, Wang T, Yuan X, et al (2018) Neural Models for Key Phrase Extraction and Question
Generation. In: Proceedings of the Workshop on Machine Reading for Question Answering. Associa-
tion for Computational Linguistics, Stroudsburg, PA, USA, pp 78–88
186. Sun X, Liu J, Lyu Y, et al (2018) Answer-focused and Position-aware Neural Question Generation. In:
Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. Associa-
tion for Computational Linguistics, Stroudsburg, PA, USA, pp 3930–3939
187. Sundermeyer M, Alkhouli T, Wuebker J, Ney H (2014) Translation Modeling with Bidirectional
Recurrent Neural Networks Human Language Technology and Pattern Recognition Group. In:
Emnlp2014. ACL, pp 14–25
188. Sutskever I, Vinyals O, Le QV (2014) Sequence to Sequence Learning with Neural Networks. Adv
Neural Inf Process Syst 4:3104–3112
189. Tambwekar P, Dhuliawala M, Martin LJ, et al (2019) Controllable Neural Story Plot Generation via
Reward Shaping. In: Proceedings of the Twenty-Eighth International Joint Conference on Artificial
Intelligence. International Joint Conferences on Artificial Intelligence Organization, California, pp
5982–5988
190. Tian C, Wang Y, Cheng H, et al (2020) Train Once, and Decode As You Like. In: Proceedings of the
28th International Conference on Computational Linguistics. International Committee on Computa-
tional Linguistics, Stroudsburg, PA, USA, pp 280–293
191. Tian Z, Yan R, Mou L, et al (2017) How to make context more useful? An empirical study on con-
text-Aware neural conversational models. ACL 2017 - 55th Annu Meet Assoc Comput Linguist Proc
Conf (Long Pap 2:231–236. https://​doi.​org/​10.​18653/​v1/​P17-​2036
192. Tu Z, Lu Z, Yang L, et al (2016) Modeling coverage for neural machine translation. 54th Annu Meet
Assoc Comput Linguist ACL 2016 - Long Pap 1:76–85. https://​doi.​org/​10.​18653/​v1/​p16-​1008
193. Upadhya BA, Udupa S, Kamath SS (2019) Deep Neural Network Models for Question Classification
in Community Question-Answering Forums. 2019 10th Int Conf Comput Commun Netw Technol
ICCCNT 2019 6–11. https://​doi.​org/​10.​1109/​ICCCN​T45670.​2019.​89448​61

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43142 Multimedia Tools and Applications (2023) 82:43089–43144

194. Vasisht S, Tirthani V, Eppa A, et al (2022) Automatic FAQ Generation Using Text-to-Text Trans-
former Model. 2022 3rd Int Conf Emerg Technol INCET 2022 1–7. https://​doi.​org/​10.​1109/​INCET​
54531.​2022.​98239​67
195. Vaswani A, Shazeer N, Parmar N, et al (2017) Attention Is All You Need. In: NIPS’17: Proceedings
of the 31st International Conference on Neural Information Processing Systems, arXiv:​1706.​03762​
v5. CA, USEA
196. Vedantam R, Zitnick CL, Parikh D (2015) CIDEr: Consensus-based image description evaluation. In:
2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, pp 4566–4575
197. Vijayakumar AK, Cogswell M, Selvaraju RR, et al (2018) Diverse Beam Search: Decoding Diverse
Solutions from Neural Sequence Models. In: 32nd AAAI Conference on Artificial Intelligence, AAAI
2018. pp 1–16
198. Vijayakumar AK, Cogswell M, Selvaraju RR, et al (2016) Diverse Beam Search: Decoding Diverse
Solutions from Neural Sequence Models. In: 32nd AAAI Conference on Artificial Intelligence, AAAI
2018. pp 7371–7379
199. Vinyals O, Toshev A, Bengio S, Erhan D (2015) Show and tell: A neural image caption generator. In:
2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, pp 3156–3164
200. Wang Q, Li B, Xiao T, et al (2019) Learning Deep Transformer Models for Machine Translation. In:
Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. Associa-
tion for Computational Linguistics, Stroudsburg, PA, USA, pp 1810–1822
201. Wang C, Yang H, Bartz C, Meinel C (2016) Image captioning with deep bidirectional LSTMs. MM
2016 - Proc 2016 ACM Multimed Conf 988–997. https://​doi.​org/​10.​1145/​29642​84.​29642​99
202. Wang P, Yang A, Men R, et al (2022) OFA: Unifying Architectures, Tasks, and Modalities Through a
Simple Sequence-to-Sequence Learning Framework. arXiv. http://​arxiv.​org/​abs/​2202.​03052
203. Wang W, Yang N, Wei F, et al (2017) Gated Self-Matching Networks for Reading Comprehension
and Question Answering. In: Proceedings of the 55th Annual Meeting of the Association for Com-
putational Linguistics (Volume 1: Long Papers). Association for Computational Linguistics, Strouds-
burg, PA, USA, pp 189–198
204. Wang T, Yuan X, Trischler A (2017) A Joint Model for Question Answering and Question Genera-
tion. arXiv:170601450v1
205. Welleck S, Kulikov I, Kim J, et al (2020) Consistency of a recurrent language model with respect
to incomplete decoding. EMNLP 2020 - 2020 Conf Empir Methods Nat Lang Process Proc Conf
5553–5568.https://​doi.​org/​10.​18653/​v1/​2020.​emnlp-​main.​448
206. Weston J, Chopra S, Bordes A (2015) Memory Networks. 3rd Int Conf Learn Represent ICLR 2015 -
Conf Track Proc 1–15
207. Wilt C, Thayer J, Ruml W (2010) A comparison of greedy search algorithms. In: Proceedings of the
3rd Annual Symposium on Combinatorial Search, SoCS 2010. SoCS 2010, pp 129–136
208. Wiseman S, Shieber SM, Rush AM (2018) Learning Neural Templates for Text Generation. Proc
2018 Conf Empir Methods Nat Lang Process EMNLP 2018 3174–3187. https://​doi.​org/​10.​18653/​v1/​
d18-​1356
209. Wolf T, Sanh V, Chaumond J, Delangue C (2019) TransferTransfo: A Transfer Learning Approach for
Neural Network Based Conversational Agents. arXiv
210. Wołk K, Koržinek D (2017) Comparison and adaptation of automatic evaluation metrics for quality
assessment of re-speaking. Comput Sci 18:129–144. https://​doi.​org/​10.​7494/​csci.​2017.​18.2.​129
211. Woodsend K, Lapata M (2010) Automatic generation of story highlights. In: Proceedings of the 48th
Annual Meeting of the Association for Computational Linguistics. Association for Computational
Linguistics, pp 565–574
212. Wu Y, Hu B (2018) Learning to Extract Coherent Summary via Deep Reinforcement Learning. In:
The Thirty-Second AAAI Conference on Artificial Intelligence (AAAI-18) Learning. Association for
the Advancement of Artificial Intelligence, pp 5602–5609
213. Wu J, Ouyang L, Ziegler DM, et al (2021) Recursively Summarizing Books with Human Feedback
214. Wu Y, Schuster M, Chen Z, et al (2016) Google’s Neural Machine Translation System: Bridging the
Gap between Human and Machine Translation. arXiv:160908144 1–23
215. Xiao Y, Wu L, Guo J, et al (2022) A Survey on Non-Autoregressive Generation for Neural Machine
Translation and Beyond. arXiv 00:1–25. http://​arxiv.​org/​abs/​2204.​09269
216. Xie Z (2017) Neural Text Generation: A Practical Guide. http://​arxiv.​org/​abs/​18030​7133 1–21
217. Xie Y, Le L, Zhou Y, Raghavan VV (2018) Deep Learning for Natural Language Processing. Handb
Stat 38:317–328. https://​doi.​org/​10.​1016/​bs.​host.​2018.​05.​001
218. Xing C, Wu W, Wu Y, et al (2017) Topic aware neural response generation. 31st AAAI Conf Artif
Intell AAAI 2017 3351–3357

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Multimedia Tools and Applications (2023) 82:43089–43144 43143

219. Xiong C, Merity S, Socher R (2016) Dynamic memory networks for visual and textual question
answering. 33rd Int Conf Mach Learn ICML 2016 5:3574–3583
220. Xu K, Ba JL, Kiros R, et al (2015) Show, Attend and Tell: Neural Image Caption Generation with
Visual Attention. In: International Conference on Machine Learning. JMLR: W&CP
221. Xu W, Li C, Lee M, Zhang C (2020) Multi-task learning for abstractive text summarization with
key information guide network. EURASIP J Adv Signal Process 2020:16. https://​doi.​org/​10.​1186/​
s13634-​020-​00674-7
222. Yamada K, Knight K (2001) A syntax-based statistical translation model. 523–530.https://​doi.​org/​
10.​3115/​10730​12.​10730​79
223. Yan Z, Duan N, Bao J, et al (2016) DocChat: An information retrieval approach for chatbot
engines using unstructured documents. 54th Annu Meet Assoc Comput Linguist ACL 2016 - Long
Pap 1:516–525. https://​doi.​org/​10.​18653/​v1/​p16-​1049
224. Yang Q, Huo Z, Shen D, et al (2019) An End-to-End Generative Architecture for Paraphrase
Generation. In: Proceedings of the 2019 Conference on Empirical Methods in Natural Language
Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-
IJCNLP). Association for Computational Linguistics, Stroudsburg, PA, USA, pp 3130–3140
225. Yang W, Xie Y, Lin A, et al (2019) End-to-End Open-Domain Question Answering with BERT-
serini. https://​doi.​org/​10.​18653/​v1/​N19-​4013
226. Yao T, Pan Y, Li Y, Mei T (2017) Incorporating copying mechanism in image captioning for
learning novel objects. Proc - 30th IEEE ConfComput Vis Pattern Recognition, CVPR 2017 2017-
Janua:5263–5271. https://​doi.​org/​10.​1109/​CVPR.​2017.​559
227. Yao L, Peng N, Weischedel R, et al (2019) Plan-and-Write: Towards Better Automatic Storytell-
ing. In: Proceedings of the AAAI Conference on Artificial Intelligence. AAAI, pp 7378–7385
228. Yao K, Zweig G, Peng B (2015) Attention with Intention for a Neural Network Conversation
Model. ArXiv, abs/151008565 1–7
229. Yin C, Qian B, Wei J, et al (2019) Automatic Generation of Medical Imaging Diagnostic Report
with Hierarchical Recurrent Neural Network. In: 2019 IEEE International Conference on Data
Mining (ICDM). IEEE, pp 728–737
230. You Q, Jin H, Wang Z, et al (2016) Image captioning with semantic attention. Proc IEEE Comput-
Soc Conf Comput Vis Pattern Recognit 2016-Decem:4651–4659. https://​doi.​org/​10.​1109/​CVPR.​
2016.​503
231. Yu L, Zhang W, Wang J, Yu Y (2017) SeqGAN : Sequence Generative Adversarial Nets with
Policy Gradient. In: 31st AAAI Conference on Artificial Intelligence. AAAI, pp 2852–2858
232. Yu W, Zhu C, Li Z et al (2022) A Survey of Knowledge-Enhanced Text Generation. ACM Comput
Surv 1:1–44. https://​doi.​org/​10.​1145/​35124​67
233. Yuan X, Wang T, Gulcehre C, et al (2017) Machine comprehension by text-to-text neural question
generation. arXiv 15–25. https://​doi.​org/​10.​18653/​v1/​w17-​2603
234. Zeiler MD (2012) ADADELTA: An Adaptive Learning Rate Method. ArXiv: 12125701
235. Zhang S, Dinan E, Urbanek J, et al (2018) Personalizing dialogue agents: I have a dog, do you
have pets too? ACL 2018 - 56th Annu Meet Assoc Comput Linguist Proc Conf (Long Pap 1:2204–
2213. https://​doi.​org/​10.​18653/​v1/​p18-​1205
236. Zhang T, Kishore V, Wu F, et al (2020) BERTScore: Evaluating Text Generation with BERT.
arXiv:190409675 1–41
237. Zhang X, Lapata M (2017) Sentence Simplification with Deep Reinforcement Learning. In: Pro-
ceedings of the 2017 Conference on Empirical Methods in Natural Language Processing. Associa-
tion for Computational Linguistics, Stroudsburg, PA, USA, pp 584–594
238. Zhang L, Sung F, Liu F, et al (2017) Actor-Critic Sequence Training for Image Captioning. ArXiv:
170609601
239. Zhang J, Tan J, Wan X (2018) Towards a Neural Network Approach to Abstractive Multi-Docu-
ment Summarization. arXiv:180107704
240. Zhang J, Zhao Y, Saleh M, Liu PJ (2019) PEGASUS: Pre-training with Extracted Gap-sentences
for Abstractive Summarization. 37th IntConf Mach Learn ICML 2020 PartF16814:11265–11276
241. Zhao Y, Ni X, Ding Y, Ke Q (2018) Paragraph-level neural question generation with maxout
pointer and gated self-attention networks. Proc 2018 Conf Empir Methods Nat Lang Process
EMNLP 2018 3901–3910. https://​doi.​org/​10.​18653/​v1/​d18-​1424
242. Zhou H, Huang M, Zhang T, et al (2018) Emotional chatting machine: Emotional conversation
generation with internal and external memory. 32nd AAAI Conf Artif Intell AAAI 2018 730–738
243. Zhou H, Huang M, Zhang T, et al (2018) Emotional chatting machine: Emotional conversation
generation with internal and external memory. Thirty-Second AAAI Conf Artif Intell 730–738

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
43144 Multimedia Tools and Applications (2023) 82:43089–43144

244. Zhou X, Wang WY (2018) MOJITALK: Generating Emotional Responses at Scale. In: Proceedings
of the 56th Annual Meeting ofthe Association for Computational Linguistics (Long Papers). Assoc
Comput Ling. Melbourne, Australia, pp 1128–1137
245. Zhou Q, Yang N, Wei F, et al (2017) Neural Question Generation from Text: A Preliminary Study.
arXiv:170401792v3 [csCL]

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under
a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted
manuscript version of this article is solely governed by the terms of such publishing agreement and applicable
law.

13
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center
GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers
and authorised users (“Users”), for small-scale personal, non-commercial use provided that all
copyright, trade and service marks and other proprietary notices are maintained. By accessing,
sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of
use (“Terms”). For these purposes, Springer Nature considers academic use (by researchers and
students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and
conditions, a relevant site licence or a personal subscription. These Terms will prevail over any
conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription (to
the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of
the Creative Commons license used will apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may
also use these personal data internally within ResearchGate and Springer Nature and as agreed share
it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not otherwise
disclose your personal data outside the ResearchGate or the Springer Nature group of companies
unless we have your permission as detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial
use, it is important to note that Users may not:

1. use such content for the purpose of providing other users with access on a regular or large scale
basis or as a means to circumvent access control;
2. use such content where to do so would be considered a criminal or statutory offence in any
jurisdiction, or gives rise to civil liability, or is otherwise unlawful;
3. falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association
unless explicitly agreed to by Springer Nature in writing;
4. use bots or other automated methods to access the content or redirect messages
5. override any security feature or exclusionary protocol; or
6. share the content in order to create substitute for Springer Nature products or services or a
systematic database of Springer Nature journal content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a
product or service that creates revenue, royalties, rent or income from our content or its inclusion as
part of a paid for service or for other commercial gain. Springer Nature journal content cannot be
used for inter-library loans and librarians may not upload Springer Nature journal content on a large
scale into their, or any other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not
obligated to publish any information or content on this website and may remove it or features or
functionality at our sole discretion, at any time with or without notice. Springer Nature may revoke
this licence to you at any time and remove access to any copies of the Springer Nature journal content
which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or
guarantees to Users, either express or implied with respect to the Springer nature journal content and
all parties disclaim and waive any implied warranties or warranties imposed by law, including
merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published
by Springer Nature that may be licensed from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a
regular basis or in any other manner not expressly permitted by these Terms, please contact Springer
Nature at

[email protected]