International Journal of Scientific Research in Computer Science, Engineering

and Information Technology

ISSN : 2456-3307 Available Online at : www.ijsrcseit.com
doi : https://doi.org/10.32628/CSEIT24103204

A Literature Review : Enhancing Sentiment Analysis of Deep

Learning Techniques Using Generative AI Model
Sharma Vishalkumar Sureshbhai1, Dr. Tulsidas Nakrani2
1Research Scholar, Department of Computer Application (MCA), Sankalchand Patel University, Visnagar, Gujarat, India
2Associate Professor, Department of Computer Application (MCA), Sankalchand Patel University, Visnagar, Guajrat, India

ARTICLEINFO ABSTRACT

Sentiment analysis is possibly one of the most desirable areas of study within Natural
Article History:
Language Processing (NLP). Generative AI can be used in sentiment analysis through
Accepted : 20 May 2024 the generation of text that reflects the sentiment or emotional tone of a given input.
Published : 15 June 2024 The process typically involves training a generative AI model on a large dataset of
text examples labeled with sentiments (positive, negative, neutral, etc.). Once
trained, the model can generate new text based on the learned patterns, providing
Publication Issue an automated way to analyze sentiments in user reviews, comments, or any other
Volume 10, Issue 3 form of textual data. The main goal of this research topic is to identify the emotions
May-June-2024 as well as opinions of users or customers using textual means. Though a lot of
research has been done in this area using a variety of models, sentiment analysis is
Page Number still regarded as a difficult topic with a lot of unresolved issues. Slang terms, novel
530-540 languages, grammatical and spelling errors, etc. are some of the current issues. This
work aims to conduct a review of the literature by utilizing multiple deep learning
methods on a range of data sets. Nearly 21 contributions, covering a variety of
sentimental analysis applications, are surveyed in the current literature study.
Initially, the analysis looks at the kinds of deep learning algorithms that are being
utilized and tries to show the contributions of each work. Additionally, the research
focuses on identifying the kind of data that was used. Additionally, each work's
performance metrics and setting are assessed, and the conclusion includes
appropriate research gaps and challenges. This will help in identifying the non-
saturated application for which sentimental analysis is most needed in future studies.
Keywords :- Sentiment Analysis, Generative AI, Deep Learning Technique,
Research Gaps and Challenges

Nomenclature

Abbreviations Descriptions
NLP Natural Language Processing

Copyright © 2024 The Author(s): This is an open-access article distributed under the terms of the Creative 530
Commons Attribution 4.0 International License (CC BY-NC 4.0)
 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

BERT Bidirectional Encoder representations from transformers

LLM Large Language Model
LSTM Long Short Term Memory
GAN Generative Adversarial Network
CNN Convolutional Neural Network
RNN Recurrent Neural Network
VAE Variational Autoencoder
GPT Generative Pre-trained Transformer
BiLSTM Bidirectional Long Short-Term Memory
EEG Electroencephalogram
IPSO Improved Particle Swarm Optimization
ASTE Aspect Sentiment Triplet Extraction
MATLAB Matrix Laboratory

I. INTRODUCTION architecture, have further pushed the boundaries of

what is achievable in sentiment analysis. Generative
Sentiment analysis, also known as opinion mining, is a models like the [18] Generative Pre-trained
natural language processing (NLP) task that involves Transformer (GPT) and[12] Bidirectional Encoder
identifying and extracting subjective information from Representations from Transformers (BERT), LSTM,
text. This process aims to determine the attitude or GAN[13], LLM, CNN, and RNN have shown
emotional tone of the writer, whether it is positive, remarkable proficiency in understanding and
negative, or neutral. The advent of deep learning and generating human-like text. These models are pre-
generative AI models has significantly advanced the trained on vast corpora and can be fine-tuned for
field of sentiment analysis, allowing for more accurate specific tasks [9], including sentiment analysis, with
and nuanced understanding of human emotions relatively smaller datasets. The integration of
expressed in text. Sentiment analysis [5] is one of the generative AI models into sentiment analysis has
most common and challenging problems in artificial spurred extensive research and development. Studies
intelligence. It uses automated tools for detecting have demonstrated their superiority over traditional
subjective information like opinions, attitudes, and and even earlier deep learning approaches in various
feelings expressed in text and are applied across news, sentiment analysis tasks, such as sentiment
blogs, and social networks. These days [6] social media, classification, aspect-based sentiment analysis, and
blogs and networks produce huge information. People emotion detection. In practical applications, these
share their thoughts on various events and issues. This advanced models are being leveraged across industries
bulk information is used for opinion and decision to enhance customer experience, monitor brand
making in a business. The sentiment analysis [3] plays reputation, and even predict market trends [3] based
a vital role in the markets regarding the investor’s on sentiment shifts. Businesses are increasingly
emotions and their belief in the market and on a adopting these technologies to gain deeper insights
specific stock. Sentiment analysis [4] using both into consumer sentiment and to make data-driven
machine learning and deep learning algorithms is decisions.
commonly used to classify a text to its respective class.
Despite their advantages, deploying generative AI
Recent advancements in generative AI models, models for sentiment analysis presents challenges such
particularly those based on the Transformer as computational resource requirements, potential

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 531

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

biases in pre-trained data and the need for fine-tuning IEEEXplore, Springer, Association for Computing
[10] to specific domains. Future research is likely to Machinery (ACM), and arXiv, to define the Recent
focus on making these models more efficient, reducing Year 2020 between 2024 of the Literature Review
biases, and improving their interpretability. The surveyed works.
integration of deep learning techniques and generative
AI models has significantly advanced the field of In 2020 Dr. G. S. N. Murthy et al. [1] have presented a
sentiment analysis, offering powerful tools to analyze Text based Sentiment Analysis using LSTM. The
and understand human emotions [13][14] and opinions objective of their study was to address the challenges
expressed in text. As technology progresses, these of manually analyzing large textual information and
models will continue to evolve, providing even more introduces the concept of sentiment analysis as an
accurate and nuanced sentiment analysis capabilities. automated process. Their research focused on the use
of sentiment analysis, particularly emphasizing its
The main contributions of this paper are portrayed as application in extracting positive and negative
follows. opinions from text data sources such as social media
comments, survey responses, and product reviews.
• To undergo a critical review of sentiment analysis They utilized LSTM on IMDB and Amazon Product
under different applications. datasets, demonstrating a superior performance LSTM
• To carry out the detailed review of various show better performance of sentiment classification
sentiment analysis models based on the deep with higher accuracy.
learning algorithms, types of data, tools, and
different performance measures. In 2021 Putra Fissabil Muhammada et al. [2] have
• To formulate the valuable research gaps and presented a Sentiment Analysis Using Word2vec and
challenges based on the existing contributions Long Short-Term Memory (LSTM) For Indonesian
under sentiment analysis. Hotel Reviews. The objective of this paper is to
highlight the necessity of developing a sentiment
The review on sentiment analysis classification is assessment technique for hotel reviews in the
designed in the following manner: Section II specifies Indonesian language. The availability of a large volume
the literature review on conventional sentimental of reviews poses a challenge for service executives who
analysis in social media. Section III describes various need to understand the impact of reviews on their
deep learning algorithms for sentiment analysis along services. They utilized LSTM on hotel reviews found
with performance measures. The analysis on different on the Traveloka website datasets, demonstrating a
types of data used and tools for sentiment analysis is superior performance LSTM show better performance
given in Section IV. The research gaps and challenges of sentiment classification with higher accuracy.
of sentiment analysis using deep learning algorithms
shown in Section V. Section VI specify the conclusion In 2021 Priyank Sonkiyaet al. [3] has presented a Stock
of the entire paper. price prediction using BERT and GAN. the main
objective of the paper is to an ensemble of methods for
II. LITERATURE REVIEW predicting stock prices. The specific focus is on
sentiment analysis of news and headlines related to
A. Related Works Apple Inc, using a version of BERT for NLP, and the
Initially, I was Read and performed an extensive search application of a Generative Adversarial Network (GAN)
at renowned databases, such as Science Direct, to predict Apple Inc's stock price based on technical

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 532

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

indicators, stock indexes, commodities, historical performance Dataset2 show better performance of
prices, and sentiment scores. They utilized BERT and sentiment classification with higher accuracy.
GAN on Apple Inc stock datasets, demonstrating a
superior performance for best result the investors and In 2021 Akana Chandra et al. [6] have presented a
traders can decide based on the predicted price for the Sentiment Analysis using Neural Network and LSTM.
next day whether to buy a particular stock or sell it. The paper identifies sentiment analysis as a key area of
interest, emphasizing the need for efficient methods to
In 2021 Ru Yang et al. [4] have presented Using GAN- manage large volumes of data and their dependencies..
based models to sentimental analysis on imbalanced They utilized LSTM on 1.6 million tweets datasets,
datasets in education domain. The objective of the demonstrating a superior performance LSTM show
importance of gathering feedback from students better performance of sentiment classification with
engaged in online learning due to the COVID-19 higher accuracy.
pandemic. The prevalence of online education, it is
crucial to understand students' opinions through their In 2022 Ali Shariq Imran et al. [7] have presented an
reviews. The paper outlines its specific focus, which is impact of synthetic text generation for sentiment
to train machine learning and deep learning models for analysis using GAN based models. The objective of the
sentiment classification using both balanced and address the issue of data imbalance in educational
imbalanced datasets. The goal is to improve sentiment feedback datasets, specifically in the context of
classification in the context of online education by sentiment analysis of users' opinions. Students’
utilizing text generation models (CatGAN and feedback is often unstructured, allowing them to
SentiGAN) to address imbalanced datasets.. They express their thoughts to open-ended questions related
utilized GAN-SentiGAN, CatGAN on CR23k-21937 to various aspects of teaching, course, and content.
course reviews from the online learning platform They utilized GAN - CatGAN and SentiGAN on
Coursera, CR100k-107016 reviews obtained from Educational feedback datasets CR23K, CR100K,
Kaggle, AMR (Amazon game and movie reviews) demonstrating a superior performance accuracy
datasets, demonstrating a superior performance The increases for CR100K dataset show better performance
results indicate that the average growth degree for of sentiment classification with higher accuracy.
CR100k is higher than CR23k.
In 2022 Burhan BİLEN et al. [8] have presented a LSTM
In 2021 Ishaani Priyadarshini et al. [5] have presented network-based sentiment analysis for customer
a Text A novel LSTM–CNN–grid search based deep reviews. The challenges faced in sentiment analysis,
neural network for sentiment analysis. The objective of particularly the complexity of human sentiments. It
the significance of sentiment analysis in the context of introduces the use of a singular tag-plural class
the increasing amount of user-generated content on approach for binary classification, employing an LSTM
the Internet. Specifically, it emphasizes the challenges network and various machine learning models. They
and importance of comprehending hidden opinions, utilized LSTM on Electronic stores and IMDB reviews
sentiments, and emotions expressed in various forms of datasets, demonstrating superior performance LSTM
online content such as emails, tweets, reviews, and Electronic stores show better performance of
comments. They utilized LSTM and CNN on Amazon sentiment classification with higher accuracy.
reviews for sentiment analysis, and IMDB Dataset of
50K Movie Reviews datasets, demonstrating a superior In 2022 Gu stavo H. de Rosa et al. [9] have presented A
Survey on Text Generation using Generative

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 533

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

Adversarial Networks. The objective of the paper is to performance LSTM show better performance of
present a comprehensive review of recent studies and sentiment classification with Amazon-Products higher
advancements in text generation using Generative accuracy.
Adversarial Networks (GANs). It highlights the
promising aspect of adversarial learning for text In 2022 Ying Cao et al. [12] have presented a Study of
generation, emphasizing its potential to generate Sentiment Analysis Algorithms for Agricultural
"natural" language. It’s also discusses the challenges Product Reviews Based on Improved BERT Model. The
posed by the fact that GANs were originally designed objective of introduce the problem of information
for continuous data (images) rather than discrete data asymmetry between consumers and producers in the
(text). They utilized GAN on Amazon Review, Chinese context of agricultural products, specifically on
Poems, COCO Captions, WMT News datasets, Internet platforms. It proposes a solution in the form of
demonstrating a superior performance WMT News a sentiment analysis algorithm based on an improved
show better performance of sentiment classification Bidirectional Encoder Representations from
with higher accuracy. Transformers (BERT) model. The goal is to effectively
analyze after-sale reviews of agricultural products on
In 2022 Nusrat Jahan Prottasha et al. [10] have the Internet, identify emotional tendencies of
presented Transfer Learning for Sentiment Analysis consumers, and extract information to optimize and
Using BERT Based Supervised Fine-Tuning. The upgrade agricultural products. They utilized BERT on
objective of the challenges and advancements in Agricultural product reviews were obtained social
sentiment analysis within the Bangla natural language media, news websites, e-commerce websites, and
processing (NLP) domain. It highlights the difficulty in offline surveys datasets, demonstrating a superior
conducting sentiment analysis for Bangla due to a lack performance Improved BERT show better
of standardized labeled data. They utilized CNN- performance of sentiment classification with higher
BiLSTM on Facebook, Twitter, and YouTube accuracy.
comments datasets, demonstrating a superior
performance CNN-BiLSTM approach wins with show In 2022 Ms. K. Chitra et al. [13] have presented a
better performance accuracy of sentiment Penalty based Sentimental Text Generation
classification. Framework using Generative Adversarial Networks.
This paper presents a novel approach using Generative
In 2022 Amjad Iqbal et al. [11] have presented a Text Adversarial Networks (GANs) for generating
Sentiment Analysis of Consumer Reviews Using Deep sentimental text. The framework introduces a penalty
Learning. The objective of the significance of mechanism within the GAN architecture to enhance
sentiment analysis in the context of consumer reviews the model's ability to generate sentiment-aware text
and social media content. The paragraph discusses the sequences. It explores the incorporation of sentiment
increasing volume of data generated by internet and constraints during the text generation process, aiming
social media platforms, with a focus on unstructured to produce more contextually relevant and emotionally
textual data. It emphasizes the challenges associated expressive content. They utilized Generative
with handling large amounts of text and the Adversarial Network, Bi-LSTM on consumer reviews
importance of accurate polarity detection in consumer of Amazon products by Kaggle datasets, demonstrating
reviews. They utilized LSTM on Amazon-Fine-Food- a superior performance Proposed Work Bi-LSTM show
Reviews, Cell Phones and Accessories, Amazon- better performance of sentiment classification with
Products datasets, demonstrating a superior higher accuracy.

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 534

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

In 2023 Chenxi Tian et al. [14] have presented a Text emphasize the significant impact of fake news on
Dual-Encoder VAE-GAN with Spatiotemporal society and highlight the importance of detecting and
Features for Emotional EEG Data Augmentation. The addressing misinformation. They utilized LLM on
overall goal is to demonstrate that the proposed model Whatsapp bot and Twitter Data datasets, In Result This
effectively learns the distribution of raw data and methodology integrated with the whatsapp bot
generates high-quality artificial samples, leading to the showcased the better result to identify the fake news
training of a high-precision affective model. They and ensure the user that the news is authentic or not.
utilized VAE-GAN on The SJTU Emotion EEG datasets,
demonstrating a superior performance the proposed In 2023 Stanislav Chumakova et al. [18] have presented
method have an average accuracy, which is a 5% a Text Generative approach to Aspect Based Sentiment
improvement compared to the original dataset. Analysis with GPT Language Models. The objective of
the paper is to introduce the limitations of existing
In 2023 Yaowei Yue et al. [15] have presented a Deep sentiment analysis methods, specifically Aspect
Learning Short Text Sentiment Analysis Based on Sentiment Triplet Extraction (ASTE) based on
Improved Particle Swarm Optimization. The objective Bidirectional Encoder Representations from
is to introduce and describe deep learning models for Transformers (BERT). The proposes an alternative
sentiment analysis of short texts. The proposed approach using an open-domain generative method
solution involves the development of a BiLSTM-TCSA based on Generative pre-trained transformer (GPT)
model (BiLSTM combining TextCNN and Self- with few-shot and fine-tuning strategies. They utilized
Attention) that utilizes an improved particle swarm LLM on laptop and restaurant reviews , Yelp restaurant
optimization (IPSO) for hyper parameter tuning. They reviews2 and Amazon and BestBuy electronics reviews
utilized BiLSTM-TCSA on IMDB movie review dataset, datasets, In Resulting models was tested on mixed
MR English movie review dataset, the SST-2 dataset domain Russian-language automatically tagged data
from the Stanford Sentiment Treebank, the Hotel with thorough large language model (LLM) with a few-
Reviews dataset, demonstrating a superior shot approach and English data, which was only
performance Hotel Reviews show better performance automatically tagged.
of sentiment classification with higher accuracy.
In 2023 Xiang Deng et al. [19] have presented LLMs to
In 2023 Abayomi Bello et al. [16] have presented a A the Moon? Reddit Market Sentiment Analysis with
BERT Framework to Sentiment Analysis of Tweets. the Large Language Models. The objective of this paper is
objective is to address sentiment analysis challenges in to describe the challenges of market sentiment analysis
micro blogging by proposing and evaluating the on social media and present a solution using semi-
effectiveness of using BERT with different neural supervised learning with a large language model
network architectures. They utilized BERT, LSTM and (LLM).its difficulty faced by human ratters in
CNN on Tweets collected from Kaggle datasets, understanding both financial markets and social media
demonstrating a superior performance the proposed j, leading to a lack of high-quality labeled data for
method have an average accuracy show better conventional supervised learning methods. They
performance. utilized LLM on Reddit datasets, demonstrating a
superior performance LLM show better performance of
In 2023 Megharani Patil et al. [17] have presented a A sentiment classification with higher accuracy.
Novel Approach to Fake News Detection Using
Generative AI. The objective of this paper is to

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 535

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

In 2023 Sio Jurnalis Pipin et al. [20] have presented a Convolutional Neural Network [5][10][15][16][20],
Sentiment Analysis Classification of ChatGPT on GAN-Generative Adversarial Network
Twitter Big Data in Indonesia Using Fast R-CNN. The [3][4][7][9][13][14], VAE- Variational Autoencoder
research aims to analyze opinions and discussions on [14], RNN- Recurrent Neural Network, and various
social media, particularly Twitter, regarding classifiers are used.
ChatGPT's potential negative impacts. To analyze a
categorized into "Positive," "Negative," and "Neutral"
sentiments. They utilized Fast R-CNN using LLM,
CNN on Twitter Data datasets, demonstrating a
superior performance R-CNN show better
performance of sentiment classification with higher
accuracy.

In 2024 Konstantinos I. et al. [21] have presented a Text

LLMs in e-commerce: A comparative analysis of GPT Fig. 1. Taxonomy of Deep Learning Algorithms
and LLaMA models in product review evaluation. The adopted for sentiment analysis
objective of this paper is to introduce and outline a
research study focused on understanding customer B. Performance Measures
satisfaction in the context of e-commerce, particularly
through the use of Large Language Models (LLMs). The Table I shows the analysis on evaluation metrics
highlights the growth of e-commerce, the integration concerning sentiment analysis using deep learning
of automated customer service processes, and the models. In this, the measures accuracy is taken into
challenges in assessing emotions within product consideration in many of the earlier contributions. The
reviews.. They utilized Large Language Model (LLM) rarely employed performance metrics are specified in
with GPT-3.5 and LLaMA-2,, BERT and RoBERT on miscellaneous. From Table I, the accuracy measure is
Amazon and eBay datasets, demonstrating a superior considered in earlier contributions. Respectively, At
performance GPT-3.5 model accuracy show better last, it is concluded that the accuracy is taken into
performance of sentiment classification. consideration in many of the earlier contributions, and
the other metrics are considered infrequently.
III. VARIOUS DEEP LEARNING ALGORITHM FOR
SENTIMENT ANALYSIS ALONG WITH
TABLE I. ACCURANCY PERFORMANCE
PERFORMANCE MEASURES MEASURES CONCERNED FOR SENTIMENT
ANALYSIS

A. Categorization of Deep Learning Algorithms

Citations Accuracy
Dr. G. S. N. Murthy et al. [1] 85%
The classification of deep learning algorithms for Fissabil Muhammada et al. [2] 85.96%
sentiment analysis is shown in Fig.1. For different Priyank Sonkiyaet al. [3] 0.73%
contributions of sentiment analysis, the techniques like Ru Yang et al. [4] 82%
BERT-Bidirectional Encoder representations from Ishaani Priyadarshini et al. [5] 0.97%
transformers[3][12][16][21], LLM-Large Language Akana Chandra et al. [6] 87%
Ali Shariq Imran et al. [7] 82%
Model [17][18][19][20][21], LSTM-Long Short Term
Burhan BİLEN et al. [8] 90.59%
Memory [1][2][5][6][8][10][11][13][15][16], CNN-

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 536

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

Gustavo H. de Rosa et al. [9] 0.943% 107016 reviews obtained from

Nusrat Jahan Prottasha et al. [10] 94.15% Kaggle, AMR(Amazon game
Amjad Iqbal et al. [11] 0.97% and movie reviews)
Ying Cao et al. [12] 89.86% Ishaani Priyadarshini IMDB Dataset of 50K Movie
Ms. K. Chitra et al. [13] 82.6% et al. [5] Reviews
Chenxi Tian et al. [14] 97.21% Akana Chandra et al. 1.6 million tweets
Yaowei Yue et al. [15] 94.59%
93% [6]
Abayomi Bello et al. [16]
Xiang Deng et al. [19] 72% Ali Shariq Imran et al. Educational feedback datasets
Sio Jurnalis Pipin et al. [20] 94.5% [7] CR23K, CR100K
Konstantinos I. et al. [21] 64.24% Burhan BİLEN et al. Electronic stores and IMDB
[8] reviews
1. ANALYSIS ON DIFFERENT TYPES OF DATA USED AND TOOLS FOR Gustavo H. de Rosa et Amazon Review , Chinese
SENTIMENT ANALYSIS al. [9] Poems, COCO Captions ,WMT
A. Data Types
News
Nusrat Jahan Prottasha Facebook, Twitter, and
et al. [10] YouTube comments
In Table II, the analysis on different data types using
Amjad Iqbal et al. [11] Amazon-Fine-Food-Reviews,
deep learning algorithms employed for sentiment
Cell Phones and Accessories,
analysis is tabulated. The twitter data is considered in Amazon-Products
[6][10][16][17][20]. In [2] [15] [18], the data related to Ying Cao et al. [12] Agricultural product reviews
hotel is taken into consideration for evaluating the were obtained social media,
efficiency of the model. In [1] [4] [9] [11] [13] [18] [21], news websites, e-commerce
the Amazon Product review data is considered for websites, and offline surveys
Ms. K. Chitra et al. consumer reviews of Amazon
evaluating the deep learning algorithms. The social
[13] products by kaggle
media information is taken in [10][12]. The data related
Chenxi Tian et al. [14] The SJTU Emotion EEG
to Educational feedback is taken in [4][7]. In [3], the datasets
stock market review data is considered for analysis. Yaowei Yue et al. [15] IMDB movie review dataset,
The movie review data is taken into consideration for MR English movie review
experimentation in [1] [4] [5] [8] [15]. The data related dataset, the SST-2 dataset from
to camera and laptop is assumed in [11][18]. the Stanford Sentiment
Treebank, the Hotel Reviews
Abayomi Bello et al. Tweets collected from Kaggle
TABLE II. ANALYSIS ON DIFFERENT DATASET
[16]
USED FOR SENTIMENT ANALYSIS UNDER Megharani Patil et al. Whatsapp bot and Twitter
DIFFERENT CONTRIBUTIONS [17] Data
Stanislav Chumakova laptop and restaurant reviews ,
Author and Citation Dataset et al. [18] Yelp restaurant reviews2 and
Dr. G. S. N. Murthy et IMDB and Amazon Product Amazon and BestBuy
al. [1] electronics reviews
Fissabil Muhammada hotel reviews found on the Xiang Deng et al. [19] Reddit dataset
et al. [2] Traveloka website Sio Jurnalis Pipin et al. Twitter Data
Priyank Sonkiyaet al. Apple Inc stock [20]
[3] Konstantinos I. et al. Amazon and eBay
Ru Yang et al. [4] CR23k-21937 course reviews [21]
from the online learning
platform Coursera,CR100k-

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 537

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

B. Execution Tool Sentiment expression can vary greatly across

cultures, making it challenging to develop universal
The execution software employed for various kinds of sentiment analysis models. What may be considered
sentiment analysis with deep learning algorithms are positive sentiment in one culture could be interpreted
diagrammatically shown in Fig. 2. In the earlier differently in another. Bridging this gap requires
contributions, the tools such as, Python, Keras, Tensor extensive cross-cultural research and understanding.
Flow, MATLAB and Google Colab are utilized. In most In many cases, sentiment analysis involves translating
of the research works, Python software is adopted for text from multilectual languages to a common language
sentiment analysis. Therefore, it is concluded that like English for analysis. However, the quality of
Python is the best tool used for sentiment analysis with translation can significantly impact the accuracy of
deep learning algorithms and still has the possibilities sentiment analysis results, leading to potential biases
of new models to be analyzed. and errors. Multilectual languages often lack the
resources and attention afforded to dominant
languages in the field of natural language processing
(NLP). This scarcity of resources can hinder the
development of effective sentiment analysis models for
these languages.

V. CONCLUSION

The present paper has developed the review of earlier

Fig. 2. Various tools used for executing sentiment
contributions with various deep learning Generative
analysis
AI models using discrete information. The present
review has explored 21 research works that covered
IV. RESEARCH GAPS AND CHALLENGES
various implementations employed for sentiment
analysis. Initially, the assessment has concentrated on
Researching sentiment analysis on multilectual
clarifying the contribution of every task and observed
languages presents several significant challenges and
the type of deep learning algorithm utilized. The
research gaps. Here are a few key points to consider:
evaluation also focused in recognizing the type of data
employed. Later, the environment utilized and the
One of the primary challenges is the availability of
performance metrics covered in each contribution was
labeled data for sentiment analysis in multilectual
analyzed. Finally, the research gaps and challenges
languages. Many sentiment analysis datasets are
were mentioned that were useful for recognizing the
primarily focused on English, leaving a gap in resources
researching sentiment analysis on multilectual
for languages spoken by diverse populations.
languages implementation for which the sentiment
Multilectual languages often exhibit significant
analysis was required in further researches.
variations in dialects, idioms, and cultural nuances.
Analyzing sentiment in such languages requires
VI. REFERENCES
models that can account for these complexities, which
may not be adequately addressed in existing techniques
[1]. Dr. G. S. N. Murthy, Shanmukha Rao Allu,
designed for more standardized languages.
Bhargavi Andhavarapu, Mounika Bagadi,
Mounika Belusonti, Text based Sentiment

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 538

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

Analysis using LSTM, International Journal of Transfer Learning for Sentiment Analysis Using
Engineering Research & Technology (IJERT) BERT Based Supervised Fine-Tuning, Sensors
ISSN: 2278-0181,Vol. 9 Issue 05, May-2020 2022, 22, 4157. https://doi.org/10.3390/s22114157
[2]. Putra Fissabil Muhammada, Retno [11]. Amjad Iqbal , Rashid Amin , Javed Iqbal , Roobaea
Kusumaningruma, Adi Wibowoa, Sentiment Alroobaea , Ahmed Binmahfoudh and Mudassar
Analysis Using Word2vec And Long Short-Term Hussain, Sentiment Analysis of Consumer
Memory (LSTM) For Indonesian Hotel Reviews, Reviews Using Deep Learning, Sustainability
Procedia Computer Science 179 (2021) 728–735 2022, 14, 10844.
[3]. Priyank Sonkiya,Vikas Bajpai,Anukriti Bansal, https://doi.org/10.3390/su141710844
Stock price prediction using BERT and GAN, [12]. Ying Cao , Zhexing Sun , Ling Li and Weinan Mo,
arXiv:2107.09055v1 [q-fin.ST] 18 Jul 2021 A Study of Sentiment Analysis Algorithms for
[4]. Ru Yang, Maryam Edalati, Using GAN-based Agricultural Product Reviews Based on Improved
models to sentimental analysis on imbalanced BERT Model, Symmetry 2022, 14, 1604.
datasets in education domain, arXiv:2108.12061v1 https://doi.org/10.3390/sym14081604
[cs.CL] 26 Aug 2021 [13]. Ms. K. Chitra, Dr. G. Kavitha, Dr. P. Latchoumy,
[5]. Ishaani Priyadarshini, Chase Cotton, A novel Penalty based Sentimental Text Generation
LSTM–CNN–grid search based deep neural Framework using Generative Adversarial
network for sentiment analysis, The Journal of Networks, International Conference on
Supercomputing (2021) 77:13911–13932 Automation, Computing and Renewable Systems
https://doi.org/10.1007/s11227-021-03838-w (ICACRS) , 978-1-6654-6084-2/22/$31.00 ©2022
[6]. Akana Chandra Mouli Venkata Srinivas, IEEE
Ch.Satyanarayana, Ch.Divakar,Katikireddy Phani https:109/ICACRS55517.2022.10029135//doi: 10.1
Sirisha, Sentiment Analysis using Neural Network [14]. Chenxi Tian, Yuliang Ma , Jared Cammon , Feng
and LSTM, IOP Conf. Series: Materials Science Fang, Yingchun Zhang , and Ming Meng, Dual-
and Engineering 1074 (2021) 012007 Encoder VAE-GAN With Spatiotemporal
https://doi:10.1088/1757-899X/1074/1/012007 Features for Emotional EEG Data Augmentation,
[7]. Ali Shariq Imran , Ru Yang , Zenun Kastrati , Sher IEEE TRANSACTIONS ON NEURAL SYSTEMS
Muhammad Daudpota , Sarang Shaikh, The AND REHABILITATION ENGINEERING, VOL.
impact of synthetic text generation for sentiment 31, 2023
analysis using GAN based models, Egyptian [15]. Yaowei Yue , Yun Peng ,and Duancheng Wang,
Informatics Journal 23 (2022) 547–557 Deep Learning Short Text Sentiment Analysis
[8]. Bilen B., Horasan F., “LSTM network based Based on Improved Particle Swarm Optimization,
sentiment analysis for customer reviews”, Electronics 2023, 12, 4119.
Politeknik Dergisi, 25(3): 959-966, (2022). DOI: https://doi.org/10.3390/electronics12194119
10.2339/politeknik.844019 [16]. Abayomi Bello, Sin-Chun Ng and Man-Fai Leung,
[9]. Gustavo H. de Rosa , Jo ao P. Papa, A Survey on A BERT Framework to Sentiment Analysis of
Text Generation using Generative Adversarial Tweets, Sensors 2023, 23, 506.
Networks, arXiv:2212.11119v1 [cs.CL] 20 Dec https://doi.org/10.3390/s23010506
2022 [17]. Megharani Patil, Hrishikesh Yadav, Mahendra
[10]. Nusrat Jahan Prottasha , Abdullah As Sami , Md Gawali, Jaya Suryawanshi, Jaikumar Patil, Anjali
Kowsher , Saydul Akbar Murad ,Anupam Kumar Yeole, Prathik Shetty, Jayesh Potlabattini, A
Bairagi , Mehedi Masud and Mohammed Baz, Novel Approach to Fake News Detection Using

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 539

 Sharma Vishalkumar Sureshbhai et al Int. J. Sci. Res. Comput. Sci. Eng. Inf. Technol., May-June-2024, 10 (3) : 530-540

Generative AI,International Journal of Intelligent

Systems and Applications in Engineering IJISAE,
2024, 12(4s), 343–354
[18]. Stanislav Chumakova,, Anton Kovantseva,
Anatoliy Surikov, Generative approach to Aspect
Based Sentiment Analysis with GPT Language
Models, Procedia Computer Science 229 (2023)
284–293
[19]. Xiang Deng,Vasilisa Bashlovkina,Feng Han,Simon
Baumgartner,Michael Bendersky, LLMs to the
Moon? Reddit Market Sentiment Analysis with
Large Language Models, WWW ’23 Companion,
April 30–May 04, 2023, Austin, TX, USA
[20]. Sio Jurnalis Pipin1, Frans Mikael Sinaga, Sunaryo
Winardi, Muhammad Noor Hakim, Sentiment
Analysis Classification of ChatGPT on Twitter Big
Data in Indonesia Using Fast R-CNN, JURNAL
MEDIA INFORMATIKA BUDIDARMA Volume
7, Nomor 4, Oktober 2023, Page 2137-2148
https://DOI: 10.30865/mib.v7i4.6816
[21]. Konstantinos I. Roumeliotis , Nikolaos D. Tselikas
, Dimitrios K. Nasiopoulos, LLMs in e-commerce:
A comparative analysis of GPT and LLaMA
models in product review evaluation, Natural
Language Processing Journal 6 (2024) 100056

Volume 10, Issue 3, May-June-2024 | http://ijsrcseit.com 540