See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/360286065

A Survey on Text Classification: From Traditional to Deep

Learning

Article in ACM Transactions on Intelligent Systems and Technology · April 2022

DOI: 10.1145/3495162

CITATIONS READS

402 3,169

8 authors, including:

Qian Li Hao Peng

Beihang University Beihang University
31 PUBLICATIONS 786 CITATIONS 279 PUBLICATIONS 7,385 CITATIONS

SEE PROFILE SEE PROFILE

Jianxin Li Congying Xia

Beihang University University of Illinois Chicago
195 PUBLICATIONS 6,564 CITATIONS 51 PUBLICATIONS 1,695 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Qian Li on 20 July 2022.

The user has requested enhancement of the downloaded file.

A Survey on Text Classification: From Traditional
to Deep Learning

QIAN LI, HAO PENG, and JIANXIN LI, Beihang University, China
CONGYING XIA, University of Illinois at Chicago, USA
RENYU YANG, University of Leeds, UK
LICHAO SUN, Lehigh University, USA
PHILIP S. YU, University of Illinois at Chicago, USA
LIFANG HE, Lehigh University, USA

Text classification is the most fundamental and essential task in natural language processing. The last decade
has seen a surge of research in this area due to the unprecedented success of deep learning. Numerous meth-
ods, datasets, and evaluation metrics have been proposed in the literature, raising the need for a compre-
hensive and updated survey. This paper fills the gap by reviewing the state-of-the-art approaches from 1961
to 2021, focusing on models from traditional models to deep learning. We create a taxonomy for text clas-
sification according to the text involved and the models used for feature extraction and classification. We
then discuss each of these categories in detail, dealing with both the technical developments and benchmark
datasets that support tests of predictions. A comprehensive comparison between different techniques, as well
as identifying the pros and cons of various evaluation metrics are also provided in this survey. Finally, we
conclude by summarizing key implications, future research directions, and the challenges facing the research
area.

CCS Concepts: • General and reference → Document types; Surveys and overviews

Additional Key Words and Phrases: Deep learning, traditional models, text classification, evaluation metrics,
challenges

ACM Reference format:

Qian Li, Hao Peng, Jianxin Li, Congying Xia, Renyu Yang, Lichao Sun, Philip S. Yu, and Lifang He. 2022.
A Survey on Text Classification: From Traditional to Deep Learning. ACM Trans. Intell. Syst. Technol. 13, 2,
Article 31 (April 2022), 41 pages.
https://doi.org/10.1145/3495162

The authors of this paper were supported by the National Key R&D Program of China through grant 2021YFB1714800,
NSFC through grants (No. U20B2053 and 61872022), State Key Laboratory of Software Development Environment (SKLSDE-
2020ZX-12). Philip S. Yu was supported by NSF under grants III-1763325, III-1909323, III-2106758, and SaTC-1930941. Lifang
He was supported by NSF ONR N00014-18-1-2009 and Lehigh’s accelerator grant S00010293. This work was also sponsored
by CAAI-Huawei MindSpore Open Fund.
Authors’ addresses: Q. Li, H. Peng, and J. Li (corresponding author), Beihang University, Haidian district, Beijing, China,
100191; emails: {liqian, penghao, lijx}@act.buaa.edu.cn; C. Xia and P. S. Yu, University of Illinois at Chicago, 601 S Morgan,
Chicago, IL, USA, 61820; emails: {cxia8, psyu}@uic.edu; R. Yang, University of Leeds, Leeds LS2 9JT, Leeds, England, UK;
email: [email protected]; L. Sun and L. He, Lehigh University, 27 Memorial Drive West, Bethlehem, PA, USA, 18015;
emails: [email protected], [email protected].
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee
provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and
the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored.
31
Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires
prior specific permission and/or a fee. Request permissions from [email protected].
© 2022 Association for Computing Machinery.
2157-6904/2022/04-ART31 $15.00
https://doi.org/10.1145/3495162

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:2 Q. Li et al.

1 INTRODUCTION
Text classification – the procedure of designating pre-defined labels for text – is an essential and
significant task in many Natural Language Processing (NLP) applications, such as sentiment
analysis [1, 2], topic labeling [3, 4], question answering [5, 6] and dialog act classification [7]. In the
era of information explosion, it is time-consuming and challenging to process and classify large
amounts of text data manually. Besides, the accuracy of manual text classification can be easily
influenced by human factors, such as fatigue and expertise. It is desirable to use machine learning
methods to automate the text classification procedure to yield more reliable and less subjective
results. Moreover, this can also help enhance information retrieval efficiency and alleviate the
problem of information overload by locating the required information.
Figure 1 illustrates a flowchart of the procedures involved in the text classification, under the
light of traditional and deep analysis. Text data is different from numerical, image, or signal data.
It requires NLP techniques to be processed carefully. The first important step is to preprocess text
data for the model. Traditional models usually need to obtain good sample features by artificial
methods and then classify them with classic machine learning algorithms. Therefore, the effective-
ness of the method is largely restricted by feature extraction. However, different from traditional
models, deep learning integrates feature engineering into the model fitting process by learning a
set of nonlinear transformations that serve to map features directly to outputs.
From the 1960s until the 2010s, traditional text classification models dominated. Traditional
methods mean statistics-based models, such as Naïve Bayes (NB) [8], K-Nearest Neighbor
(KNN) [9], and Support Vector Machine (SVM) [10]. Comparing with the earlier rule-based
methods, this method has obvious advantages in accuracy and stability. However, these approaches
still need to do feature engineering, which is time-consuming and costly. Besides, they usually dis-
regard the natural sequential structure or contextual information in textual data, making it chal-
lenging to learn the semantic information of the words. Since the 2010s, text classification has
gradually changed from traditional models to deep learning models. Compared with the methods
based on traditional, deep learning methods avoid designing rules and features by humans and au-
tomatically provide semantically meaningful representations for text mining. Therefore, most of
the text classification research works are based on Deep Neural Networks (DNNs) [11], which
are data-driven approaches with high computational complexity. Few works focus on traditional
models to settle the limitations of computation and data.

1.1 Major Differences and Contributions

There have been several works reviewing text classification and its subproblems recently. Two of
them are reviews of text classification. Kowsari et al. [12] surveyed different text feature extraction,
dimensionality reduction methods, basic model structure for text classification, and evaluation
methods. Minaee et al. [13] reviewed recent deep learning based text classification methods, bench-
mark datasets, and evaluation metrics. Unlike existing text classification reviews, we conclude ex-
isting models from traditional models to deep learning with works of recent years. Traditional
models emphasize the feature extraction and classifier design. Once the text has well-designed
characteristics, it can be quickly converged by training the classifier. DNNs can perform feature
extraction automatically and learn well without domain knowledge. We then give the datasets and
evaluation metrics for single-label and multi-label tasks and summarize future research challenges
from data, models, and performance perspectives. Moreover, we summarize various information
in three tables, including the necessary information of classic deep learning models, primary in-
formation of main datasets, and a general benchmark of state-of-the-art methods under different
applications. In summary, this study’s main contributions are as follows:

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:3

Fig. 1. Flowchart of the text classification with classic methods in each module. It is crucial to extract essen-
tial features for traditional methods, but features can be extracted automatically by deep learning methods.

• We introduce the process and development of text classification and present comprehensive
analysis and research on primary models – from traditional to deep learning models – ac-
cording to their model structures. We summarize the necessary information of deep learning
models in terms of basic model structures in Table 1, including publishing years, methods,
venues, applications, evaluation metrics, datasets, and code links.
• We introduce the present datasets and give the formulation of main evaluation metrics with
the comparison of metrics, including single-label and multi-label text classification tasks. We
summarize the necessary information of primary datasets in Table 2, including the number
of categories, average sentence length, the size of each dataset, related papers, and data
addresses.
• We summarize classification accuracy scores of models given in their articles, on benchmark
datasets in Table 4 and conclude the survey by discussing the main challenges facing the text
classification and key implications stemming from this study.

1.2 Organization of the Survey

The rest of the survey is organized as follows. Section 2 summarizes the existing models related
to text classification, including traditional and deep learning models, including a summary table.
Section 3 introduces the primary datasets with a summary table and evaluation metrics on single-
label and multi-label tasks. We then give quantitative results of the leading models in classic text
classification datasets in Section 4. Finally, we summarize the main challenges for deep learning
text classification in Section 5 before concluding the article in Section 6.

2 TEXT CLASSIFICATION METHODS

Text classification is referred to as extracting features from raw text data and predicting the cat-
egories of text data based on such features. Numerous models have been proposed in the past
few decades for text classification. For traditional models, NB [8] is the first model used for the
text classification task. Whereafter, generic classification models are proposed, such as KNN [9],
SVM [10], and Random Forest (RF) [14], which are called classifiers, and are widely used for text
classification. Recently, the eXtreme Gradient Boosting (XGBoost) [15] and the Light Gradi-
ent Boosting Machine (LightGBM) [16] have arguably the potential to provide excellent per-
formance. For deep learning models, TextCNN [17] has the highest number of references in these
models, wherein a Convolutional Neural Network (CNN) [18] model has been introduced to
solve the text classification problem for the first time. While not specifically designed for han-
dling text classification tasks, the Bidirectional Encoder Representation from Transformers
(BERT) [19] has been widely employed when designing text classification models, considering its
effectiveness on numerous text classification datasets.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:4 Q. Li et al.

2.1 Traditional Models

Traditional models accelerate text classification with improved accuracy and make the application
scope of traditional expand. The first thing is to preprocess the raw input text for training
traditional models, which generally consists of word segmentation, data cleaning, and statistics.
Then, text representation aims to express preprocessed text in a form that is much easier for
computers and minimizes information loss, such as Bag-Of-Words (BOW) [20], N-gram [21],
Term Frequency-Inverse Document Frequency (TF-IDF) [22], word2vec [23], and Global
Vectors for word representation (GloVe) [24]. BOW means that all the words in the corpus are
formed into a mapping array. According to the mapping array, a sentence can be represented
as a vector. The i-th element in the vector represents the frequency of the i-th word in the
mapping array of the sentence. The vector is the BOW of the sentence. At the core of the BOW is
representing each text with a dictionary-sized vector. The individual value of the vector denotes
the word frequency corresponding to its inherent position in the text. Compared to BOW, N-gram
considers the information of adjacent words and builds a dictionary by considering the adjacent
words. It is used to calculate the probability model of a sentence. The probability of a sentence is
expressed as the joint probability of each word in the sentence. The probability of a sentence can
be calculated by predicting the probability of the N -th word, given the sequence of the (N − 1)-th
words. To simplify the calculation, the N-gram model adopts the Markov hypothesis [21]. A word
appears only concerning the words that preceded it. Therefore, the N-gram model performs a
sliding window with size N. By counting and recording the occurrence frequency of all fragments,
the probability of a sentence can be calculated using the frequency of relevant fragments in the
record. TF-IDF [22] uses the word frequency and inverses the document frequency to model the
text. TF is the word frequency of a word in a specific article, and IDF is the reciprocal of the
proportion of the articles containing this word to the total number of articles in the corpus. TF-IDF
is the multiplication of the two. TF-IDF assesses the importance of a word to one document in a set
of files or a corpus. The importance of a word increases proportionally with the number of times it
appears in a document. However, it decreases inversely with its frequency in the corpus as a whole.
The word2vec [23] employs local context information to obtain word vectors, as shown in Figure 2.
Word vector refers to a fixed-length real value vector specified as the word vector for any word in
the corpus. The word2vec uses two essential models: CBOW and Skip-gram. The former is to pre-
dict the current word on the premise that the context of the current word is known. The latter is to
predict the context when the current word is known. The GloVe [24] – with both the local context
and global statistical features – trains on the nonzero elements in a word-word co-occurrence
matrix, as shown in Figure 2. It enables word vectors to contain as much semantic and grammatical
information as possible. The construction method of word vector is: firstly, the co-occurrence
matrix of words is constructed based on the corpus, and then the word vector is learned based on
the co-occurrence matrix and GloVe model. Finally, the represented text is fed into the classifier
according to selected features. Here, we discuss some representative classifiers in detail.

2.1.1 PGM-based Methods. Probabilistic Graphical Models (PGMs) express the conditional
dependencies among features in graphs, such as the Bayesian network [25]. It is a combination of
probability theory and graph theory.
Naïve Bayes (NB) [8] is the simplest and most broadly used model based on applying Bayes’
theorem. The NB algorithm has an independent assumption: when the target value has been given,
the conditions between text T = [T1 ,T2 , . . . ,Tn ] are independent (see Figure 3). The NB algorithm
primarily

uses the prior probability to calculate the posterior probability P (y | T1 ,T2 , . . . ,Tn ) =
p (y ) nj=1 p (T j |y )
n . Due to its simple structure, NB is broadly used for text classification tasks.
j =1 p (T j )

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:5

Fig. 2. The structure of word2vec, including CBOW and Skip-gram.

Fig. 3. The structure of Naïve Bayes.

Fig. 4. The structure of KNN where k = 4 (left) and the structure of SVM (right). Each node represents a text
and nodes with different contours represent different categories.

Although the assumption that the features are independent is sometimes not actual, it substan-
tially simplifies the calculation process and performs better. To improve the performance on
smaller categories, Schneider [26] proposes a feature selection score method through calculating
KL-divergence [27] between the training set and corresponding categories for multinomial NB text
classification. Dai et al. [28] propose a transfer learning method named Naive Bayes Transfer
Classification (NBTC) to settle the different distribution between the training set and the target
set. It uses the EM algorithm [29] to obtain a locally optimal posterior hypothesis on the target set.
NB classifier is also used for fake news detection [30], and sentiment analysis [31], which can be
seen as a text classification task. Bernoulli NB, Gaussian NB and Multinomial NB are three popular
approaches of NB text classification [32]. Multinomial NB performs slightly better than Bernoulli
NB on few labeled datasets [33]. Bayesian NB classifier with Gaussian event model [32] has been
proven to be superior to NB with multinomial event model on 20 Newsgroups (20NG) [34] and
WebKB [35] datasets.
2.1.2 KNN-based Methods. At the core of the K-Nearest Neighbors (KNN) algorithm [9] is to
classify an unlabeled sample by finding the category with the most samples on the k nearest sam-
ples. It is a simple classifier without building the model and can decrease complexity through
the fasting process of getting k nearest neighbors. Figure 4 showcases the structure of KNN.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:6 Q. Li et al.

We can find k training texts approaching a specific text to be classified through estimating the
in-between distance. Hence, the text can be divided into the most common categories found in k
training set texts. The improvement of KNN algorithm mainly includes feature similarity [36], K
value [37] and index optimization [38]. However, due to the positive correlation between model
time/space complexity and the amount of data, the KNN algorithm takes an unusually long time
on the large-scale datasets [39]. To decrease the number of selected features, Soucy et al. [40] pro-
pose a KNN algorithm without feature weighting. It manages to find relevant features, building
the inter-dependencies of words by using a feature selection. When the data is extremely unevenly
distributed, KNN tends to classify samples with more data. The Neighbor-Weighted K-Nearest
Neighbor (NWKNN) [41] is proposed to improve classification performance on the unbalanced
corpora. It casts a significant weight for neighbors in a small category and a small weight for
neighbors in a broad class.
2.1.3 SVM-based Methods. Cortes and Vapnik [42] propose Support Vector Machine (SVM)
to tackle the binary classification of pattern recognition. Joachims [10], for the first time, uses the
SVM method for text classification representing each text as a vector. As illustrated in Figure 4,
SVM-based approaches turn text classification tasks into multiple binary classification tasks. In
this context, SVM constructs an optimal hyperplane in the one-dimensional input space or feature
space, maximizing the distance between the hyperplane and the two categories of training sets,
thereby achieving the best generalization ability. The goal is to make the distance of the category
boundary along the direction perpendicular to the hyperplane to be the largest. Equivalently, this
will result in the lowest error rate of classification. Constructing an optimal hyperplane can be
transformed into a quadratic programming problem to obtain a globally optimal solution. Choos-
ing the appropriate kernel function [43] and feature selection [44] are of the utmost importance
to ensure SVM can deal with nonlinear problems and become a robust nonlinear classifier. Fur-
thermore, active learning [45] and adaptive learning [46] method are used for text classification to
reduce the labeling effort based on the supervised learning algorithm SVM. To analyze what the
SVM algorithms learn and what tasks are suitable, Joachims [47] proposes a theoretical learning
model combining the statistical traits with the generalization performance of an SVM, analyzing
the features and benefits using a quantitative approach. Transductive Support Vector Machine
(TSVM) [48] is proposed to lessen misclassifications of the particular test collections with a gen-
eral decision function considering a specific test set. It uses prior knowledge to establish a more
suitable structure and study faster.
2.1.4 DT-based Methods. Decision Trees (DT) [49] is a supervised tree structure learning
method – reflective of the idea of divide-and-conquer – and is constructed recursively. It learns
disjunctive expressions and has robustness for the text with noise. As shown in Figure 5, deci-
sion trees can be generally divided into two distinct stages: tree construction and tree pruning. It
starts at the root node and tests the data samples (composed of instance sets, which have several
attributes), and divides the dataset into diverse subsets according to different results. A subset of
datasets constitutes a child node, and every leaf node in the decision tree represents a category.
Constructing the decision tree is to determine the correlation between classes and attributes, fur-
ther exploited to predict the record categories of unknown forthcoming types. The classification
rules generated by the decision tree algorithm are straight-forward, and the pruning strategy [50]
can also help reduce the influence of noise. Its limitation, however, mainly derives from inefficiency
in coping with explosively increasing data size. More specifically, the Iterative Dichotomiser 3
(ID3) [51] algorithm uses information gain as the attribute selection criterion in the selection of
each node – It is used to select the attribute of each branch node, and then select the attribute
having the maximum information gain value to become the discriminant attribute of the current

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:7

Fig. 5. An example of DT (left) and the structure of RF (right). The nodes with the dotted outline represent
the nodes of the decision route. It has five features to predict whether each text belongs to category A or B.

node. Based on ID3, C4.5 [52] learns to obtain a map from attributes to classes, which effectively
classifies entities unknown to new categories. DT based algorithms usually need to train for each
dataset, which is low efficiency [53]. Thus, Johnson et al. [54] propose a DT-based symbolic rule
system. The method represents each text as a vector calculated by the frequency of each word in
the text, and induces rules from the training data. The learning rules are used for classifying the
other data, being similar to the training data. Furthermore, to reduce the computational costs of DT
algorithms, Fast Decision-Tree (FDT) [55] uses a two-pronged strategy: pre-selecting a feature
set and training multiple DTs on different data subsets. Results from multiple DTs are combined
through a data-fusion technique to resolve the cases of imbalanced classes.

2.1.5 Integration-based Methods. Integrated algorithms aim to aggregate the results of multi-
ple algorithms for better performance and interpretation. Conventional integrated algorithms are
bootstrap aggregation, such as RF [14], boosting such as the Adaptive Boosting (AdaBoost)
[56] and XGBoost [15], and stacking. The bootstrap aggregation method trains multiple classifiers
without strong dependencies and then aggregates their results. For instance, RF [14] consists of
multiple tree classifiers wherein all trees depend on the value of the random vector sampled inde-
pendently (depicted in Figure 5). It is worth noting that each tree within the RF shares the same
distribution. The generalization error of an RF relies on the strength of each tree and the relation-
ship among trees, and will converge to a limit with the increment of tree number in the forest. In
boosting based algorithms, all labeled data are trained with the same weight to initially obtain a
weaker classifier [57]. The weights of the data will then be adjusted according to the former result
of the classifier. The training procedure will continue by repeating such steps until the termination
condition is reached. Unlike bootstrap and boosting algorithms, stacking based algorithms break
down the data into n parts and use n classifiers to calculate the input data in a cascade manner –
Results from the upstream classifier will feed into the downstream classifier as input. The training
will terminate once a pre-defined iteration number is targeted. The integrated method can cap-
ture more features from multiple trees. However, it helps little for short text. Motivated by this,
Bouaziz et al. [58] combine data enrichment – with semantics in RFs for short text classification –
to overcome the deficiency of sparseness and insufficiency of contextual information. In integrated
algorithms, not all classifiers learn well. It is necessary to give different weights for each classifier.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:8 Q. Li et al.

To differentiate contributions of trees in a forest, Islam et al. [59] exploit the Semantics Aware
Random Forest (SARF) classifier, choosing features similar to the features of the same class, for
extracting features and producing the prediction values.
Summary. The parameters of NB are more diminutive, less sensitive to missing data, and the
algorithm is simple. However, it assumes that features are independent of each other. When the
number of features is large, or the correlation between features is significant, the performance
of NB decreases. SVM can solve high-dimensional and nonlinear problems. It has a high general-
ization ability, but it is sensitive to missing data. KNN mainly depends on the surrounding finite
adjacent samples, rather than discriminating class domain to determine the category. Thus, for
the dataset to be divided with more crossover or overlap of the class domain, it is more suitable
than other methods. DT is easy to understand and interpret. Given an observed model, it is easy
to deduce the corresponding logical expression according to the generated decision tree. The tra-
ditional method is a type of machine learning. It learns from data, which are pre-defined features
that are important to the performance of prediction values. However, feature engineering is tough
work. Before training the classifier, we need to collect knowledge or experience to extract features
from the original text. The traditional methods train the initial classifier based on various textual
features extracted from the raw text. Toward small datasets, traditional models usually present
better performance than deep learning models under the limitation of computational complexity.
Therefore, some researchers have studied the design of traditional models for specific domains
with less data.

2.2 Deep Learning Models

The DNNs consist of artificial neural networks that simulate the human brain to automatically
learn high-level features from data, getting better results than traditional models in speech recog-
nition, image processing, and text understanding. Input datasets should be analyzed to classify the
data, such as a single-label, multi-label, unsupervised, unbalanced dataset. According to the trait
of the dataset, the input word vectors are sent into the DNN for training until the termination
condition is reached. The performance of the training model is verified by the downstream task,
such as sentiment classification, question answering, and event prediction. We show some DNNs
over the years in Table 1, including designs that are different from the corresponding basic models,
evaluation metrics, and experimental datasets.
Numerous deep learning models have been proposed in the past few decades for text classifica-
tion, as shown in Table 1. We tabulate primary information – including publication years, venues,
applications, code links, evaluation metrics, and experiment datasets – of main deep learning mod-
els for text classification. The applications in this table include Sentiment Analysis (SA), Topic
Labeling (TL), News Classification (NC), Question Answering (QA), Dialog Act Classifi-
cation (DAC), Natural Language Inference (NLI), and Relation Classification (RC). The
multilayer perceptron [172] and the recursive neural network [173] are the first two deep learning
approaches used for the text classification task, which improve performance compared with tra-
ditional models. Then, CNNs, Recurrent Neural Networks (RNNs), and attention mechanisms
are used for text classification [101, 174, 175]. Many researchers advance text classification perfor-
mance for different tasks by improving CNN, RNN, and attention, or model fusion and multi-task
methods. The appearance of BERT [19], which can generate contextualized word vectors, is a sig-
nificant turning point in the development of text classification and other NLP technologies. Many
researchers [142, 176] have studied text classification models based on BERT, which achieves better
performance than the above models in multiple NLP tasks, including text classification. Besides,
some researchers study text classification technology based on Graph Neural Network (GNN)

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:9

Table 1. Basic Information Based on Different Models

Model Year Method Venue Applications Code Link Metrics Datasets
2011 RAE [60] EMNLP SA, QA [61] Accuracy MPQA, MR, EP
ReNN 2012 MV-RNN [62] EMNLP SA [63] Accuracy, F1 MR
2013 RNTN [64] EMNLP SA [65] Accuracy SST
2014 DeepRNN [66] NIPS SA, QA - Accuracy SST-1;SST-2
MLP 2014 Paragraph-Vec [67] ICML SA, QA [68] Error Rate SST, IMDB
2015 DAN [69] ACL SA, QA [70] Accuracy, Time RT, SST, IMDB
2015 Tree-LSTM [1] ACL SA [71] Accuracy SST-1, SST-2
2015 S-LSTM [2] ICML SA - Accuracy SST
2015 TextRCNN [72] AAAI SA, TL [73] Macro-F1, etc. 20NG, Fudan, ACL, SST-2
2015 MT-LSTM [6] EMNLP SA,QA [74] Accuracy SST-1, SST-2, QC, IMDB
2016 oh-2LSTMp [75] ICML SA, TL [76] Error Rate IMDB, Elec, RCV1, 20NG
RNN 2016 BLSTM-2DCNN [77] COLING SA, QA, TL [78] Accuracy SST-1, Subj, TREC, etc.
2016 Multi-Task [79] IJCAI SA [80] Accuracy SST-1, SST-2, Subj, IMDB
2017 DeepMoji [81] EMNLP SA [82] Accuracy SS-Twitter, SE1604, etc.
2017 TopicRNN [83] ICML SA [84] Error Rate IMDB
2017 Miyato et al. [85] ICLR SA [86] Error Rate IMDB, DBpedia, etc.
2018 RNN-Capsule [87] TheWebConf SA [88] Accuracy MR, SST-1, etc.
2019 HM-DenseRNNs [89] IJCAI SA, TL [90] Accuracy IMDB, SST-5, AG
2014 TextCNN [17] EMNLP SA, QA [91] Accuracy MR, SST-2, Subj, etc.
2014 DCNN [5] ACL SA, QA [92] Accuracy MR, TREC, Twitter
2015 CharCNN [93] NeurIPS SA, QA, TL [94] Error Rate AG, Yelp P, DBPedia, etc.
2016 SeqTextRCNN [7] NAACL Dialog act [95] Accuracy DSTC 4, MRDA, SwDA
2017 XML-CNN [96] SIGIR NC, TL, SA [97] NDCG@K, etc. EUR-Lex, Wiki-30K, etc.
CNN 2017 DPCNN [98] ACL SA, TL [99] Error Rate AG, DBPedia, Yelp.P, etc.
2017 KPCNN [100] IJCAI SA, QA, TL - Accuracy Twitter, AG, Bing, etc.
2018 TextCapsule [101] EMNLP SA, QA, TL [102] Accuracy Subj, TREC, Reuters, etc.
2018 HFT-CNN [103] EMNLP TL [104] Micro-F1, etc. RCV1, Amazon670K
2019 CCRCNN [105] AAAI TL - Accuracy TREC, MR, AG
2020 Bao et al. [106] ICLR TL [107] Accuracy 20NG, Reuters-2157, etc.
2016 HAN [108] NAACL SA, TL [109] Accuracy Yelp.F, YahooA, etc.
2016 BI-Attention [110] NAACL SA - Accuracy NLP&CC 2013 [111]
2016 LSTMN [112] EMNLP SA [113] Accuracy SST-1
2017 Lin et al. [114] ICLR SA [115] Accuracy Yelp, SNLI Age
2018 SGM [116] COLING TL [117] HL, Micro-F1 RCV1-V2, AAPD
2018 ELMo [118] NAACL SA, QA, NLI [119] Accuracy SQuAD, SNLI, SST-5
Attention 2018 BiBloSA [120] ICLR SA [121] Accuracy, Time CR, MPQA, SUBJ, etc.
2019 AttentionXML [122] NeurIPS TL [123] P@k, N@k, etc. EUR-Lex, etc.
2019 HAPN [124] EMNLP RC - Accuracy FewRel, CSID
2019 Proto-HATT [125] AAAI RC [126] Accuracy FewRel
2019 STCKA [127] AAAI SA, TL [128] Accuracy Weibo, Product Review, etc.
2020 HyperGAT [129] EMNLP TL, NC [130] Accuracy 20NG, Ohsumed, MR, etc.
2020 MSMSA [131] AAAI ST, QA, NLI - Accuracy, F1 IMDB, MR, SST, SNLI, etc.
2020 Choi [132] EMNLP SA, TL - Accuracy SST2, IMDB, 20NG
2019 BERT [19] NAACL SA, QA [133] Accuracy SST-2, QQP, QNLI, CoLA
2019 BERT-BASE [134] ACL TL [135] P@K, R@K, etc. EUR-LEX
2019 Sun et al. [136] CCL SA, QA, TL [137] Error Rate TREC, DBPedia, etc.
2019 XLNet [138] NeurIPS SA, QA, NC [139] EM, F1, etc. Yelp-2, AG, MNLI, etc.
2019 RoBERTa [140] arXiv SA, QA [141] F1, Accuracy SQuAD, MNLI-m, SST-2
Trans 2020 GAN-BERT [142] ACL SA, NLI [143] F1, Accuracy SST-5, MNLI
2020 BAE [144] EMNLP SA, QA [145] Accuracy Amazon, Yelp, MR, MPQA
2020 ALBERT [146] ICLR SA, QA [147] F1, Accuracy SST, MNLI, SQuAD
2020 TG-Transformer [148] EMNLP SA, TL - Accuracy, Time R8, R52, Ohsumed, etc.
2020 X-Transformer [149] KDD SA, TL [150] P@K, R@K Eurlex-4K, Wiki10-31K, etc.
2021 LightXML [151] arXiv TL, ML, NLI [152] P@K, Time AmazonCat-13K, etc.
2018 DGCNN [153] TheWebConf TL [154] Macro-F1, etc. RCV1, NYTimes
2019 TextGCN [155] AAAI SA, TL [156] Accuracy 20NG, Ohsumed, R52, etc.
2019 SGC[157] ICML NC, TL, SA [158] Accuracy, Time 20NG, R8, Ohsumed, etc.
GNN 2019 Huang et al. [159] EMNLP NC, TL [160] Accuracy R8, R52, Ohsumed
2019 Peng et al. [161] arXiv NC, TL - Micro-F1, etc. RCV1, EUR-Lex, etc.
2020 TextING [162] ACL SA, NC, TL [163] Accuracy MR, R8, R52, Ohsumed
2020 TensorGCN [164] AAAI SA, NC, TL [165] Accuracy 20NG, R8, R52, Ohsumed, MR
2020 MAGNET [166] ICAART TL [167] Micro-F1, HL Reuters, RCV1-V2, etc.
2017 Miyato et al. [85] ICLR SA, NC [168] Error Rate IMDB, RCV1, et al.
Others 2018 TMN [169] EMNLP TL - Accuracy, F1 Snippets, Twitter, et al.
2019 Zhang et al. [170] NAACL TL, NC [171] Accuracy DBpedia, 20NG.
Trans: Transformer. Time: training time.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:10 Q. Li et al.

Fig. 6. The architecture of ReNN (left) and the architecture of MLP (right).

[155, 177] to capture structural information in the text, which cannot be replaced by other methods.
Here, we classify DNNs by structure and discuss some representative models in detail.

2.2.1 ReNN-based Methods. Traditional models cost lots of time on design features for each
task. Furthermore, in the case of deep learning, the meaning of “word vectors” is different: each
input word is associated with a fixed-length vector whose values are either drawn at random or
derived from a previous traditional process, thus forming a matrix L called word embedding matrix
which represents the vocabulary words in a small latent semantic space, of generally 50 to 300 di-
mensions. The Recursive Neural Network (ReNN) [173] can automatically learn the semantics
of text recursively and the syntax tree structure without feature design, as shown in Figure 6. We
give an example of ReNN based models. First, each word of input text is taken as the leaf node of
the model structure. Then all nodes are combined into parent nodes using a weight matrix. The
weight matrix is shared across the whole model. Each parent node has the same dimension with all
leaf nodes. Finally, all nodes are recursively aggregated into a parent node to represent the input
text to predict the label.
ReNN-based models improve performance compared with traditional models and save on la-
bor costs due to excluding feature designs used for different text classification tasks. The Recur-
sive AutoEncoder (RAE) [60] is used to predict the distribution of sentiment labels for each in-
put sentence and learn the representations of multi-word phrases. To learn compositional vector
representations for each input text, the Matrix-Vector Recursive Neural Network (MV-RNN)
[62] introduces a ReNN model to learn the representation of phrases and sentences. It allows that
the length and type of input texts are inconsistent. MV-RNN allocates a matrix and a vector for
each node on the constructed parse tree. Furthermore, the Recursive Neural Tensor Network
(RNTN) [64] is proposed with a tree structure to capture the semantics of sentences. It inputs
phrases with different lengths and represents the phrases by parse trees and word vectors. The
vectors of higher nodes on the parse tree are estimated by the equal tensor-based composition
function. For RNTN, the time complexity of building the textual tree is high, and expressing the
relationship between documents is complicated within a tree structure. The performance is usually
improved, with the depth being increased for DNNs. Therefore, Irsoy et al. [66] propose a Deep
Recursive Neural Network (DeepReNN), which stacks multiple recursive layers. It is built by
binary parse trees and learns distinct perspectives of compositionality in language.

2.2.2 MLP-based Methods. A MultiLayer Perceptron (MLP) [172], sometimes colloquially

called “vanilla” neural network, is a simple neural network structure that is used for capturing
features automatically. As shown in Figure 6, we show a three-layer MLP model. It contains an
input layer, a hidden layer with an activation function in all nodes, and an output layer. Each node

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:11

Fig. 7. The RNN based model (left) and the CNN based model (right).

connects with a certain weight w i . It treats each input text as a bag of words and achieves high
performance on many text classification benchmarks comparing with traditional models.
There are some MLP-based methods proposed by some research groups for text classification
tasks. The Paragraph Vector (Paragraph-Vec) [67] is the most popular and widely used method,
which is similar to the Continuous Bag-Of-Words (CBOW) [23]. It gets fixed-length feature
representations of texts with various input lengths by employing unsupervised algorithms. Com-
paring with CBOW, it adds a paragraph token mapped to the paragraph vector by a matrix. The
model predicts the fourth word by the connection or average of this vector to the three contexts
of the word. Paragraph vectors can be used as a memory for paragraph themes and are used as a
paragraph function and inserted into the prediction classifier.

2.2.3 RNN-based Methods. The Recurrent Neural Network (RNN) [173] is broadly used
for capturing long-range dependency through recurrent computation. The RNN language model
learns historical information, considering the location information among all words suitable for
text classification tasks. We show an RNN model for text classification with a simple sample, as
shown in Figure 7. Firstly, each input word is represented by a specific vector using a word embed-
ding technology. Then, the embedding word vectors are fed into RNN cells one by one. The output
of RNN cells are the same dimension with the input vector and are fed into the next hidden layer.
The RNN shares parameters across different parts of the model and has the same weights of each
input word. Finally, the label of input text can be predicted by the last output of the hidden layer.
To diminish the time complexity of the model and capture contextual information, Liu et al. [79]
introduce a model for catching the semantics of long texts. It is a biased model that parsed the text
one by one, making the following inputs profit over the former and decreasing the semantic effi-
ciency of capturing the whole text. For modeling topic labeling tasks with long input sequences,
TopicRNN [83] is proposed. It captures the dependencies of words in a document via latent topics
and uses RNNs to capture local dependencies and latent topic models for capturing global seman-
tic dependencies. Virtual Adversarial Training (VAT) [178] is a useful regularization method
applicable to semi-supervised learning tasks. Miyato et al. [85] apply adversarial and virtual ad-
versarial training text and employ the perturbation into word embedding rather than the original
input text. The model improves the quality of the word embedding and is not easy to overfit dur-
ing training. Capsule network [179] captures the relationships between features using dynamic
routing between capsules comprised of a group of neurons in a layer. Wang et al. [87] propose an
RNN-Capsule model with a simple capsule structure for the sentiment classification task.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:12 Q. Li et al.

In the backpropagation process of RNN, the weights are adjusted by gradients, calculated by
continuous multiplications of derivatives. If the derivatives are extremely small, it may cause a
gradient vanishing problem by continuous multiplications. Long Short-Term Memory (LSTM)
[180], the improvement of RNN, effectively alleviates the gradient vanishing problem. It is com-
posed of a cell to remember values on arbitrary time intervals and three gate structures to control
information flow. The gate structures include input gates, forget gates, and output gates. The LSTM
classification method can better capture the connection among context feature words, and use the
forgotten gate structure to filter useless information, which is conducive to improving the total
capturing ability of the classifier. Tree-LSTM [1] extends the sequence of LSTM models to the tree
structure. The whole subtree with little influence on the result can be forgotten through the LSTM
forgetting gate mechanism for the Tree-LSTM model.
Natural Language Inference (NLI) [181] predicts whether one text’s meaning can be deduced
from another by measuring the semantic similarity between each pair of sentences. To consider
other granular matchings and matchings in the reverse direction, Wang et al. [182] propose a
model for the NLI task named Bilateral Multi-Perspective Matching (BiMPM). It encodes input
sentences by the BiLSTM encoder. Then, the encoded sentences are matched in two directions.
The results are aggregated in a fixed-length matching vector by another BiLSTM layer. Finally, the
result is evaluated by a fully connected layer.

2.2.4 CNN-based Methods. Convolutional Neural Networks (CNNs) [18] are proposed for
image classification with convolving filters that can extract features of pictures. Unlike RNN, CNN
can simultaneously apply convolutions defined by different kernels to multiple chunks of a se-
quence. Therefore, CNNs are used for many NLP tasks, including text classification. For text clas-
sification, the text requires being represented as a vector similar to the image representation, and
text features can be filtered from multiple angles, as shown in Figure 7. Firstly, the word vectors
of the input text are spliced into a matrix. The matrix is then fed into the convolutional layer,
which contains several filters with different dimensions. Finally, the result of the convolutional
layer goes through the pooling layer and concatenates the pooling result to obtain the final vector
representation of the text. The category is predicted by the final vector.
To try using CNN for the text classification task, an unbiased model of convolutional neural
networks is introduced by Kim, called TextCNN [17]. It can better determine discriminative phrases
in the max-pooling layer with one layer of convolution and learn hyperparameters except for
word vectors by keeping word vectors static. Training only on labeled data is not enough for data-
driven deep models. Therefore, some researchers consider utilizing unlabeled data. Johnson et al.
[183] propose a CNN model based on two-view semi-supervised learning for text classification,
which first uses unlabeled data to train the embedding of text regions and then labeled data. DNNs
usually have better performance, but it increases the computational complexity. Motivated by this,
a Deep Pyramid Convolutional Neural Network (DPCNN) [98] is proposed, with a little more
computational accuracy, increasing by raising the network depth. The DPCNN is more specific
than Residual Network (ResNet) [184], as all the shortcuts are exactly simple identity mappings
without any complication for dimension matching.
According to the minimum embedding unit of text, embedding methods are divided into
character-level, word-level, and sentence-level embedding. Character-level embeddings can settle
Out-Of-Vocabulary (OOV) [185] words. Word-level embeddings learn the syntax and seman-
tics of the words. Moreover, sentence-level embedding can capture relationships among sentences.
Motivated by these, Nguyen et al. [186] propose a deep learning method based on a dictionary,
increasing information for word-level embeddings through constructing semantic rules and deep
CNN for character-level embeddings. Adams et al. [187] propose a character-level CNN model,

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:13

Fig. 8. The architecture of hierarchical attention network (HAN) [108].

called MGTC, to classify multi-lingual texts written. TransCap [188] is proposed to encapsulate
the sentence-level semantic representations into semantic capsules and transfer document-level
knowledge.
RNN based models capture the sequential information to learn the dependency among input
words, and CNN based models extract the relevant features from the convolution kernels. Thus
some works study the fusion of the two methods. BLSTM-2DCNN [77] integrates a Bidirectional
LSTM (BiLSTM) with two-dimensional max pooling. It uses a 2D convolution to sample more
meaningful information of the matrix and understands the context better through BiLSTM. More-
over, Xue et al. [189] propose MTNA, a combination of BiLSTM and CNN layers, to solve aspect
category classification and aspect term extraction tasks.
2.2.5 Attention-based Methods. CNN and RNN provide excellent results on tasks related to text
classification. However, these models are not intuitive enough for poor interpretability, especially
in classification errors, which cannot be explained due to the non-readability of hidden data. The
attention-based methods are successfully used in the text classification. Bahdanau et al. [190] first
propose an attention mechanism that can be used in machine translation. Motivated by this, Yang
et al. [108] introduce the Hierarchical Attention Network (HAN) to gain better visualization by
employing the extremely informational components of a text, as shown in Figure 8. HAN includes
two encoders and two levels of attention layers. The attention mechanism lets the model pay
different attention to specific inputs. It aggregates essential words into sentence vectors firstly and
then aggregates vital sentence vectors into text vectors. The attention-based models can learn how
much contribution of each word and sentence for the classification judgment, which is beneficial
for applications and analysis through the two levels of attention.
The attention mechanism can improve the performance with interpretability for text classifi-
cation, which makes it popular. There are some other works based on attention. LSTMN [112] is
proposed to process text step by step from left to right and does superficial reasoning through
memory and attention. BI-Attention [110] is designed for cross-lingual text classification to catch
bilingual long-distance dependencies. Hu et al. [191] propose an attention mechanism based on

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:14 Q. Li et al.

Fig. 9. An example of self-attention for calculating output vector b2 .

category attributes for solving the imbalance of the number of various charges which contain
few-shot charges. HAPN [124] is presented for few-shot text classification.
Self-attention [192] captures the weight distribution of words in sentences by constructing K, Q,
and V matrices among sentences that can capture long-range dependencies on text classification.
We give an example for self-attention, as shown in Figure 9. Each input word vector ai can be
represented as three n-dimensional vectors, including qi , ki, and vi . After
√ self-attention, the output

vector bi can be represented as j so f tmax (ai j )v j and ai j = qi · k j / n. All output vectors can
be parallelly computed. Lin et al. [114] used source token self-attention to explore the weight
of every token to the entire sentence in the sentence representation task. To capture long-range
dependencies, Bi-directional Block Self-Attention Network (Bi-BloSAN) [120] uses an intra-
block Self-Attention Network (SAN) to every block split by sequence and an inter-block SAN
to the outputs.
Aspect-Based Sentiment Analysis (ABSA) [31, 193] breaks down a text into multiple as-
pects and allocates each aspect a sentiment polarity. The sentiment polarity can be divided into
three types: positive, neutral, and negative. Some attention-based models are proposed to iden-
tify the fine-grained opinion polarity towards a specific aspect for aspect-based sentiment tasks.
ATAE-LSTM [194] can concentrate on different parts of each sentence according to the input
through the attention mechanisms. MGAN [195] presents a fine-grained attention mechanism
with a coarse-grained attention mechanism to learn the word-level interaction between context
and aspect.
To catch the complicated semantic relationship among each question and candidate answers for
the QA task, Tan et al. [196] introduce CNN and RNN and generate answer embeddings by using
a simple one-way attention mechanism affected through the question context. The attention cap-
tures the dependence among the embeddings of questions and answers. Extractive QA can be seen
as the text classification task. It inputs a question and multiple candidates answers and classifies
every candidate answer to recognize the correct answer. Furthermore, AP-BILSTM [197] with a
two-way attention mechanism can learn the weights between the question and each candidate
answer to obtain the importance of each candidate answer to the question.
2.2.6 Pre-trained Methods. Pre-trained language models [198] effectively learn global semantic
representation and significantly boost NLP tasks, including text classification. It generally uses un-
supervised methods to mine semantic knowledge automatically and then constructs pre-training
targets so that machines can learn to understand semantics.
As shown in Figure 10, we give differences in the model architectures among the Embedding
from Language Model (ELMo) [118], OpenAI GPT [199], and BERT [19]. ELMo [118] is a deep

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:15

Fig. 10. Differences in pre-trained model architectures [19], including BERT, OpenAI GPT, and ELMo. Ei
represents embedding of i th input. Trm represents the transformer block. Ti represents predicted tag of i th
input.

contextualized word representation model, which is readily integrated into models. It can model
complicated characteristics of words and learn different representations for various linguistic con-
texts. It learns each word embedding according to the context words with the bi-directional LSTM.
GPT [199] employs supervised fine-tuning and unsupervised pre-training to learn general repre-
sentations that transfer with limited adaptation to many NLP tasks. Furthermore, the domain of
the target dataset does not need to be similar to the domain of unlabeled datasets. The training
procedure of the GPT algorithm usually includes two stages. Firstly, the initial parameters of a
neural network model are learned by a modeling objective on the unlabeled dataset. We can then
employ the corresponding supervised objective to accommodate these parameters for the target
task. To pre-train deep bidirectional representations from the unlabeled text through joint condi-
tioning on both left and right context in every layer, the BERT model [19], proposed by Google,
significantly improves performance on NLP tasks, including text classification. BERT applies the
bi-directional encoder designed to pre-train the bi-directional representation of depth by jointly
adjusting the context in all layers. It can utilize contextual information when predicting which
words are masked. It is fine-tuned by adding just an additional output layer to construct models
for multiple NLP tasks, such as SA, QA, and machine translation. Comparing with these three
models, ELMo is a feature-based method using LSTM, and BERT and OpenAI GPT are fine-tuning
approaches using Transformer. Furthermore, ELMo and BERT are bidirectional training models
and OpenAI GPT is training from left to right. Therefore, BERT gets a better result, which com-
bines the advantages of ELMo and OpenAI GPT.
Transformer-based models can parallelize computation without considering the sequential in-
formation suitable for large scale datasets, making it popular for NLP tasks. Thus, some other
works are used for text classification tasks and get excellent performance. RoBERTa [140], is an
improved version of BERT, and adopts the dynamic masking method that generates the masking
pattern every time with a sequence to be fed into the model. It uses more data for longer pre-
training and estimates the influence of various essential hyperparameters and the size of training
data. To be specific: (1) The training time is longer (a total of nearly 200,000 training set, nearly
1.6 billion training data have been seen), the batch size (8K) is larger, and the training data is
more (30G Chinese training, including 300 million sentences and 10 billion words); (2) It removes
the next sentence prediction (NSP) task; (3) It employs more extended training sequence; 4) It
dynamically adjusts the masking mechanism and use the full word mask.
XLNet [138] is a generalized autoregressive pre-training approach. Unlike BERT, the denoising
autoencoder with the mask is not used in the first stage, but the autoregressive LM is used. It
maximizes the expected likelihood across the whole factorization order permutations to learn the
bidirectional context. Furthermore, it can overcome the weaknesses of BERT by an autoregressive
formulation and integrate ideas from Transformer-XL [200] into pre-training.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:16 Q. Li et al.

Fig. 11. The architecture of BART [202] and SpanBERT [203].

The BERT model has many parameters. In order to reduce the parameters, ALBERT [146] uses
two-parameter simplification schemes. It reduces the fragmentation vector’s length and shares pa-
rameters with all encoders. It also replaces the next sentence matching task with the next sentence
order task and continuously blocks fragmentation. When the ALBERT model is pre-trained on a
massive Chinese corpus, the parameters are fewer than BERT and also have better performance. In
general, these methods adopt unsupervised objective functions for pre-training, including the next
sentence prediction, masking technology, and permutation. These target functions based on the
word prediction demonstrate a strong ability to learn the word dependence and semantic structure
[201].
BART [202] is a denoising autoencoder based on the Seq2Seq model, as shown in Figure 11(a).
The pre-training of BART consists of two steps. Firstly, it uses a noise function to destroy the text.
Secondly, the Seq2Seq model is used to reconstruct the original text. In various noise methods,
the model obtains optimal performance by randomly shuffling the order of the original sentence
and then using the first new text filling method. The new text filling method is replacing the text
fragment with a single mask token. It uses only a specific masked token to indicate that a token is
masked.
SpanBERT [203] is specially designed to better represent and predict spans of text, as shown
in Figure 11(b). It optimizes BERT from three aspects and achieves good results in multiple tasks
such as QA. The specific optimization is embodied in three aspects. Firstly, the span mask scheme is
proposed to mask a continuous paragraph of text randomly. Secondly, Span Boundary Objective
(SBO) is added to predict span by the token next to the span boundary to get the better performance
to finetune stage. Thirdly, the NSP pre-training task is removed.
ERNIE [204] is based on the method of knowledge enhancement. It learns the semantic relations
in the real world by modeling the prior semantic knowledge such as entity concepts in massive
datasets. Specifically, ERNIE enables the model to learn the semantic representation of complete
concepts by masking semantic units such as words and entities. It mainly consists of a Trans-
former encoder and task embedding. In the Transformer encoder, the context information of each
token is captured by the self-attention mechanism, and the context representation is generated for
embedding. Task embedding is used for tasks with different characteristics.

2.2.7 GNN-based Methods. The DNN models like CNN get great performance on regular struc-
ture, but not for arbitrarily structured graphs. Some researchers study how to expand on arbitrarily
structured graphs [205, 206]. With the increasing attention of Graph Neural Networks (GNNs),
GNN-based models [207, 208] obtain excellent performance by encoding syntactic structure of sen-
tences on semantic role labeling task [209], relation classification task [210], and machine transla-
tion task [211]. It turns text classification into a graph node classification task. We show a GCN
model for text classification with four input texts, as shown in Figure 12. Firstly, the four input
ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:17

Fig. 12. The GNN-based model. The different initial graph depends on how the graph is designed. We give
an example to establish edges between documents and documents, documents and sentences, and words to
words.

texts T = [T1 ,T2 ,T3 ,T4 ] and the words X = [x 1 , x 2 , x 3 , x 4 , x 5 , x 6 ] in the text, defined as nodes, are
constructed into the graph structures. The graph nodes are connected by bold black edges, which
indicates document-word edges and word-word edges. The weight of each word-word edge usu-
ally means their co-occurrence frequency in the corpus. Then, the words and texts are represented
through the hidden layer. Finally, the label of all input texts can be predicted by the graph.
The GNN-based models can learn the syntactic structure of sentences, making some researchers
study using GNN for text classification. DGCNN [153] is a graph-CNN converting text to graph-of-
words, having the advantage of learning different levels of semantics with CNN models. Yao et al.
[155] propose the Text Graph Convolutional Network (TextGCN), which builds a heteroge-
neous word text graph for a whole dataset and captures global word co-occurrence information.
To enable GNN-based models to underpin online testing, Huang et al. [159] build graphs for each
text with global parameter sharing, not a corpus-level graph structure, to help preserve global in-
formation and reduce the burden. TextING [162] builds individual graphs for each document and
learns text-level word interactions by GNN to effectively produce embeddings for obscure words
in the new text.
Graph ATtention network (GAT) [212] employs masked self-attention layers by attending
over its neighbors. Thus, some GAT-based models are proposed to compute the hidden represen-
tations of each node. The Heterogeneous Graph ATtention networks (HGAT) [213] with a
dual-level attention mechanism learns the importance of different neighboring nodes and node
types in the current node. The model propagates information on the graph and captures the rela-
tions to address the semantic sparsity for semi-supervised short text classification. MAGNET [166]
is proposed to capture the correlation among the labels based on GATs, which learns the crucial
dependencies between the labels and generates classifiers by a feature matrix and a correlation
matrix.
Event Prediction (EP) can be divided into generated event prediction and selective event pre-
diction (also known as script event prediction). EP, referring to scripted event prediction in this
review, infers the subsequent event according to the existing event context. Unlike other text clas-
sification tasks, texts in EP are composed of a series of sequential subevents. Extracting features of
the relationship among such subevents is of critical importance. SGNN [214] is proposed to model
event interactions and learn better event representations by constructing an event graph to utilize
the event network information better. The model makes full use of dense event connections for
the EP task.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:18 Q. Li et al.

2.2.8 Others. In addition to all the above models, there are some other individual models. Here
we introduce some exciting models.

Siamese Neural Network. The siamese neural network [215] is also called a twin neural net-
work (Twin NN). It utilizes equal weights while working in tandem using two distinct input
vectors to calculate comparable output vectors. Mueller et al. [216] present a siamese adaptation
of the LSTM network comprised of couples of variable-length sequences. The model is employed
to estimate the semantic similarity among texts, exceeding carefully handcrafted features and pro-
posed neural network models of higher complexity. The model further represents text employing
neural networks whose inputs are word vectors learned separately from a vast dataset. To settle
unbalanced data classification in the medical domain, Jayadeva et al. [217] use a Twin NN model
to learn from enormous unbalanced corpora. The objective functions achieve the Twin SVM ap-
proach with non-parallel decision boundaries for the corresponding classes, and decrease the Twin
NN complexity, optimizing the feature map to better discriminate among classes.
Virtual Adversarial Training (VAT). Deep learning methods require many extra hyperpa-
rameters, which increase the computational complexity. VAT [218], regularization based on local
distributional smoothness can be used in semi-supervised tasks, requires only some hyperparame-
ters, and can be interpreted directly as robust optimization. Miyato et al. [85] use VAT to effectively
improve the robustness and generalization ability of the model and word embedding performance.
Reinforcement Learning (RL). RL learns the best action in a given environment through maxi-
mizing cumulative rewards. Zhang et al. [219] offer an RL approach to establish structured sentence
representations via learning the structures related to tasks. The model has Information Distilled
LSTM (ID-LSTM) and Hierarchical Structured LSTM (HS-LSTM) representation models. The
ID-LSTM learns the sentence representation by choosing essential words relevant to tasks, and the
HS-LSTM is a two-level LSTM for modeling sentence representation.

Memory Networks. Memory networks [220] learn to combine the inference components and the
long-term memory component. Li et al. [221] use two LSTMs with extended memories and neural
memory operations for jointly handling the extraction tasks of aspects and opinions via memory
interactions. Topic Memory Networks (TMN) [169] is an end-to-end model that encodes latent
topic representations indicative of class labels.

QA Style for Sentiment Classification Task. It is an interesting attempt to treat the sentiment
classification task as a QA task. Shen et al. [222] create a high-quality annotated corpus. A three-
stage hierarchical matching network was proposed to consider the matching information between
questions and answers.

External Commonsense Knowledge. Due to the insufficient information of the event itself to dis-
tinguish the event for the EP task, Ding et al. [223] consider that the event extracted from the
original text lacked common knowledge, such as the intention and emotion of the event partici-
pants. The model improves the effect of stock prediction, EP, and so on.

Quantum Language Model. In the quantum language model, the words and dependencies among
words are represented through fundamental quantum events. Zhang et al. [224] design a quantum-
inspired sentiment representation method to learn both the semantic and the sentiment informa-
tion of subjective text. By inputting density matrices to the embedding layer, the performance of
the model improves.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:19

Summary. RNN computes sequentially and cannot be calculated in parallel. The shortcoming
of RNN makes it more challenging to become mainstream in the current trend that models tend
to have deeper and more parameters. CNN extracts features from text vectors through the con-
volution kernel. The number of features captured by the convolution kernel is related to its size.
CNN is deep enough that, in theory, it can capture features at long distances. Due to insufficient
optimization methods for parameters of the deep network and the loss of location information
due to the pooling layer, the deeper layer does not bring significant improvement. Compared with
RNN, CNN has parallel computing capability and can effectively retain location information for
the improved version of CNN. Still, it has weak feature capture capability for long-distance. GNN
builds a graph for text. When a valid graph structure is designed, the learned representation can
better capture the structural information. Transformer treats the input text as a fully connected
graph, with attention score weights on the edges. It is capable of parallel computing and is highly
efficient in extracting features between different words by self-attention, solving short-term mem-
ory problems. However, the attention mechanism in Transformer is computation-heavy, especially
when dealing with long sequences. Some improved models [146, 225] for computing complexity in
Transformer have recently been proposed. Overall, Transformer is a better choice for text classifi-
cation. Deep Learning consists of multiple hidden layers in a neural network with a higher level
of complexity and can be trained on unstructured data. Deep learning can learn language features
and master higher level and more abstract language features based on words and vectors. Deep
learning architecture can learn feature representations directly from the input without too many
manual interventions and prior knowledge. However, deep learning technology is a data-driven
method that requires enormous data to achieve high performance. Although self-attention based
models can bring some interpretability among words for DNNs, it is not enough comparing with
traditional models to explain why and how it works well.

3 DATASETS AND EVALUATION METRICS

3.1 Datasets
The availability of labeled datasets for text classification has become the main driving force behind
the fast advancement of this research field. In this section, we summarize the characteristics of
these datasets in terms of domains and give an overview in Table 2, including the number of
categories, average sentence length, the size of each dataset, related papers, data sources to access,
and applications.

3.1.1 Sentiment Analysis (SA). SA is the process of analyzing and reasoning the subjective text
within emotional color. It is crucial to get information on whether it supports a particular point of
view from the text that is distinct from the traditional text classification that analyzes the objective
content of the text. SA can be binary or multi-class. Binary SA is to divide the text into two cate-
gories, including positive and negative. Multi-class SA classifies text to multi-level or fine-grained
labels. The SA datasets include Movie Review (MR) [226, 257], Stanford Sentiment Treebank
(SST) [227], Multi-Perspective Question Answering (MPQA) [229, 258], IMDB [230], Yelp
[231], Amazon Reviews (AM) [93], NLP&CC 2013 [111], Subj [250], CR [251], SS-Twitter [259],
SS-Youtube [259], SE1604 [260], and so on. Here we detail several datasets.
MR. The MR is a movie review dataset, each of which corresponds to a sentence. The corpus
has 5,331 positive data and 5,331 negative data. 10-fold cross-validation by random splitting is
commonly used to test MR.
SST. The SST is an extension of MR. It has two categories. SST-1 with fine-grained labels with
five classes. It has 8,544 training texts and 2,210 test texts, respectively. Furthermore, SST-2 has

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:20 Q. Li et al.

Table 2. Summary Statistics for the Datasets

Datasets #C #L #N Language Related Papers Sources Applications
MR 2 20 10,662 English [5, 17, 101, 155] [226] SA
SST-1 5 18 11,855 English [17, 64] [1, 2][112] [227] SA
SST-2 2 19 9,613 English [6, 17, 64] [19, 79] [228] SA
MPQA 2 3 10,606 English [17, 60, 120] [229] SA
IMDB 2 294 50,000 English [69][108] [6] [79] [85] [230] SA
Yelp.P 2 153 598,000 English [93, 98] [231] SA
Yelp.F 5 155 700,000 English [93, 98, 108] [231] SA
Amz.P 2 91 4,000,000 English [93, 122] [232] SA
Amz.F 5 93 3,650,000 English [93, 108, 122] [232] SA
Twitter 3 19 11,209 English [5][100] [233] SA
NLP&CC 2013 2 - 115,606 Multi-language [110] [111] SA
20NG 20 221 18,846 English [72, 75, 106, 155, 157] [34] NC
AG News 4 45/7 127,600 English [98, 100] [101, 138] [234] NC
R8 8 66 7,674 English [155, 157] [159] [235] NC
R52 52 70 9,100 English [155, 157] [159] [235] NC
Sogou 6 578 510,000 Chinese [93] [236] NC
Newsgroup 20 - 18,846 English [237] [237] NC
DBPedia 14 55 630,000 English [85, 93, 98, 136] [238] TL
Ohsumed 23 136 7,400 English [155, 157, 159] [239] TL
YahooA 10 112 1,460,000 English [93, 108] [93] TL
EUR-Lex 3,956 1,239 19,314 English [96] [134, 161] [134] [240] TL
Amazon670K 670 244 643,474 English [103, 122] [241] TL
Google news 152 6 11,109 English [3, 242, 243] [242] TL
TweetSet 2011-2012 89 - 2,472 English [242, 243] [242] TL
TweetSet 2011-2015 269 8 30,322 English [3, 4] [4] TL
Bing 4 20 34,871 English [100] [244] TL
Fudan 20 2981 18,655 Chinese [72] [245] TL
SQuAD - 5,000 5,570 English [118, 118, 140, 146] [246] QA
TREC-QA - 1,162 68 English [197] [247] QA
TREC 6 10 5,952 English [5, 6, 17] [100] [248] QA
WikiQA - 873 243 English [197, 249] [249] QA
Subj 2 23 10,000 English [17, 79, 101] [250] QA
CR 2 19 3,775 English [17, 101] [251] QA
Reuters 90 168 10,788 English [101, 166] [252] ML
Reuters10 10 168 9,979 English [253] [254] ML
RCV1 103 240 807,595 English [75, 103, 134, 153] [255] ML
RCV1-V2 103 124 804,414 English [116, 166] [256] ML
AAPD 54 163 55,840 English [116, 166] [117] ML
C: Number of target classes. L: Average sentence length. N: Dataset size.

9,613 texts with binary labels being partitioned into 6,920 training texts, 872 development texts,
and 1,821 testing texts.
MPQA. The MPQA is an opinion dataset. It has two class labels and also an MPQA dataset
of opinion polarity detection sub-tasks. MPQA includes 10,606 sentences extracted from news
articles from various news sources. It should be noted that it contains 3,311 positive texts and
7,293 negative texts without labels of each text.
IMDB reviews. The IMDB review is developed for binary sentiment classification of film re-
views with the same amount in each class. It can be separated into training and test groups on
average, by 25,000 comments per group.
Yelp reviews. The Yelp review is summarized from the Yelp Dataset Challenges in 2013, 2014,
and 2015. This dataset has two categories. Yelp-2 of these were used for negative and positive

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:21

emotion classification tasks, including 560,000 training texts and 38,000 test texts. Yelp-5 is used
to detect fine-grained affective labels with 650,000 training and 50,000 test texts in all classes.
AM. The AM is a popular corpus formed by collecting Amazon website product reviews [232].
This dataset has two categories. The Amazon-2 with two classes includes 3,600,000 training sets
and 400,000 testing sets. Amazon-5, with five classes, includes 3,000,000 and 650,000 comments for
training and testing.
3.1.2 News Classification (NC). News content is one of the most crucial information sources
which has a critical influence on people. The NC system facilitates users to get vital knowledge
in real-time. News classification applications mainly encompass recognizing news topics and rec-
ommending related news according to user interest. The news classification datasets include 20
Newsgroups (20NG) [34], AG News (AG) [93, 234], R8 [235], R52 [235], Sogou News (Sogou)
[136], and so on. Here we detail several datasets.
20NG. The 20NG is a newsgroup text dataset. It has 20 categories with the same number of each
category and includes 18,846 texts.
AG. The AG News is a search engine for news from academia, choosing the four largest classes.
It uses the title and description fields of each news. AG contains 120,000 texts for training and
7,600 texts for testing.
R8 and R52. R8 and R52 are two subsets which are the subset of Reuters [252]. R8 has 8 cate-
gories, divided into 2,189 test files and 5,485 training courses. R52 has 52 categories, split into 6,532
training files and 2,568 test files.
Sogou. The Sogou combines two datasets, including SogouCA and SogouCS news sets. The
label of each text is the domain name in the URL.
3.1.3 Topic Labeling (TL). The topic analysis attempts to get the meaning of the text by defining
the sophisticated text theme. The topic labeling is one of the essential components of the topic
analysis technique, intending to assign one or more subjects for each document to simplify the
topic analysis. The topic labeling datasets include DBPedia [238], Ohsumed [239], Yahoo answers
(YahooA) [93], EUR-Lex [240], Amazon670K [241], Bing [244], Fudan [245], and PubMed [261].
Here we detail several datasets.
DBpedia. The DBpedia is a large-scale multi-lingual knowledge base generated using
Wikipedia’s most ordinarily used infoboxes. It publishes DBpedia each month, adding or delet-
ing classes and properties in every version. DBpedia’s most prevalent version has 14 classes and
is divided into 560,000 training data and 70,000 test data.
Ohsumed. The Ohsumed belongs to the MEDLINE database. It includes 7,400 texts and has 23
cardiovascular disease categories. All texts are medical abstracts and are labeled into one or more
classes.
YahooA. The YahooA is a topic labeling task with 10 classes. It includes 140,000 training data
and 5,000 test data. All texts contains three elements, being question titles, question contexts, and
best answers, respectively.
3.1.4 Question Answering (QA). The QA task can be divided into two types: the extractive QA
and the generative QA. The extractive QA gives multiple candidate answers for each question to
choose which one is the right answer. Thus, the text classification models can be used for the
extractive QA task. The QA discussed in this paper is all extractive QA. The QA system can apply
the text classification model to recognize the correct answer and set others as candidates. The
question answering datasets include Stanford Question Answering Dataset (SQuAD) [246],
TREC-QA [248], WikiQA [249], Subj [250], CR [251], MS MARCO [262], and Quora [263]. Here we
detail several datasets.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:22 Q. Li et al.

SQuAD. The SQuAD is a set of question and answer pairs obtained from Wikipedia articles. The
SQuAD has two categories. SQuAD1.1 contains 536 pairs of 107,785 Q&A items. SQuAD2.0 com-
bines 100,000 questions in SQuAD1.1 with more than 50,000 unanswerable questions that crowd
workers face in a form similar to answerable questions [264].
TREC-QA. The TREC-QA includes 5,452 training texts and 500 testing texts. It has two versions.
TREC-6 contains 6 categories, and TREC-50 has 50 categories.
WikiQA. The WikiQA dataset includes questions with no correct answer, which needs to eval-
uate the answer.
MS MARCO. The MS MARCO contains questions and answers. The questions and part of the
answers are sampled from actual web texts by the Bing search engine. Others are generative. It is
used for developing generative QA systems released by Microsoft.
3.1.5 Natural Language Inference (NLI). NLI is used to predict whether the meaning of one
text can be deduced from another. Paraphrasing is a generalized form of NLI. It uses the task
of measuring the semantic similarity of sentence pairs to decide whether one sentence is the in-
terpretation of another. The NLI datasets include Stanford Natural Language Inference (SNLI)
[181], Multi-Genre Natural Language Inference (MNLI) [265], Sentences Involving Compo-
sitional Knowledge (SICK) [266], Microsoft Research Paraphrase (MSRP) [267], Semantic
Textual Similarity (STS) [268], Recognising Textual Entailment (RTE) [269], SciTail [270],
etc. Here we detail several of the primary datasets.
SNLI. The SNLI is generally applied to NLI tasks. It contains 570,152 human-annotated sentence
pairs, including training, development, and test sets, which are annotated with three categories:
neutral, entailment, and contradiction.
MNLI. The MNLI is an expansion of SNLI, embracing a broader scope of written and spoken
text genres. It includes 433,000 sentence pairs annotated by textual entailment labels.
SICK. The SICK contains almost 10,000 English sentence pairs. It consists of neutral, entailment,
and contradictory labels.
MSRP. The MSRP consists of sentence pairs, usually for the text-similarity task. Each pair is
annotated by a binary label to discriminate whether they are paraphrases. It respectively includes
1,725 training and 4,076 test sets.
3.1.6 Multi-Label (ML) Datasets. In multi-label classification, an instance has multiple labels,
and each label can only take one of the multiple classes. There are many datasets based on multi-
label text classification. It includes Reuters [252], Reuters Corpus Volume I (RCV1) [255], RCV1-
2K [255], Arxiv Academic Paper Dataset (AAPD) [117], Patent, Web of Science (WOS-11967)
[271], AmazonCat-13K [272], BlurbGenreCollection (BGC) [273], etc. Here we detail several
datasets.
Reuters. The Reuters is a popularly used dataset for text classification from Reuters financial
news services. It has 90 training classes, 7,769 training texts, and 3,019 testing texts, containing
multiple labels and single labels. There are also some Reuters sub-sets of data, such as R8, BR52,
RCV1, and RCV1-v2.
RCV1 and RCV1-2K. The RCV1 is collected from Reuters News articles from 1996-1997, which
is human-labeled with 103 categories. It consists of 23,149 training and 784,446 testing texts, re-
spectively. The RCV1-2K dataset has the same features as the RCV1. However, the label set of
RCV1-2K has been expanded with some new labels. It contains 2,456 labels.
AAPD. The AAPD is a large dataset in the computer science field for the multi-label text classi-
fication from website.1 It has 55,840 papers, including the abstract and the corresponding subjects
1 https://arxiv.org/.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:23

Table 3. The Notations Used in Evaluation Metrics

Notations Descriptions
TP true positive
FP false positive
TN true negative
FN false negative
T Pt true positive of the t th label on a text
F Pt false positive of the t th label on a text
T Nt true negative of the t th label on a text
F Nt false negative of the t th label on a text
S label set of all samples
Q the number of predicted labels on each text

with 54 labels in total. The aim is to predict the corresponding subjects of each paper according to
the abstract.
Patent Dataset. The Patent Dataset is obtained from USPTO,2 which is a patent system grat-
ing U.S. patents containing textual details such title and abstract. It contains 100,000 US patents
awarded in the real-world with multiple hierarchical categories.
WOS-11967. The WOS-11967 is crawled from the Web of Science, consisting of abstracts of
published papers with two labels for each example. It is more shallow, but significantly broader,
with fewer classes in total.
3.1.7 Others. There are some datasets for other applications, such as SemEval-2010 Task 8 [274],
ACE 2003-2004 [275], TACRED [276], and NYT-10 [277], FewRel [278], Dialog State Tracking
Challenge 4 (DSTC 4) [279], ICSI Meeting Recorder Dialog Act (MRDA) [280], and Switch-
board Dialog Act (SwDA) [281], and so on.

3.2 Evaluation Metrics

In terms of evaluating text classification models, accuracy and F1 score are the most used to assess
the text classification methods. Later, with the increasing difficulty of classification tasks or the
existence of some particular tasks, the evaluation metrics are improved. For example, evaluation
metrics such as P@K and Micro−F 1 are used to evaluate multi-label text classification performance,
and MRR is usually used to estimate the performance of QA tasks. In Table 3, we give the notations
used in evaluation metrics.
3.2.1 Single-label Metrics. Single-label text classification divides the text into one of the most
likely categories applied in NLP tasks such as QA, SA, and dialogue systems [7]. For single-label
text classification, one text belongs to just one catalog, making it possible not to consider the
relations among labels. Here, we introduce some evaluation metrics used for single-label text clas-
sification tasks.
Accuracy and ErrorRate. The Accuracy and ErrorRate are the fundamental metrics for a text
classification model. The Accuracy and ErrorRate are respectively defined as
(T P + T N )
Accuracy = , (1)
N

(F P + F N )
ErrorRate = 1 − Accuracy = . (2)
N
2 https://www.uspto.gov/.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:24 Q. Li et al.

Precision, Recall and F1. These are vital metrics utilized for unbalanced test sets, regardless of
the standard type and error rate. For example, most of the test samples have a class label. F 1 is the
harmonic average of Precision and Recall. Precision, Recall, and F 1 as defined
TP TP
Precision = , Recall = , (3)
TP + FP TP + FN
2Precision × Recall
F1 = . (4)
Precision + Recall
The desired results will be obtained when the accuracy, F 1 and Recall value reach 1. On the con-
trary, when the values become 0, the worst result is obtained. For the multi-class classification
problem, the precision and recall value of each class can be calculated separately, and then the
performance of the individual and whole can be analyzed.
Exact Match (EM). The EM [29] is a metric for QA tasks, measuring the prediction that matches
all the ground-truth answers precisely. It is the primary metric utilized on the SQuAD dataset.
Mean Reciprocal Rank (MRR). The MRR [282] is usually applied for assessing the performance
of ranking algorithms on QA and Information Retrieval (IR) tasks. MRR is defined as
Q
1  1
MRR = , (5)
Q i=1 rank (i)
where rank (i) is the ranking of the ground-truth answer at answer i-th.
Hamming-Loss (HL). The HL [57] assesses the score of misclassified instance-label pairs where
a related label is omitted or an unrelated label is predicted.
Among these single-label evaluation metrics, the Accuracy is the earliest metric that calculates
the proportion of the sample size that is predicted correctly and is not considered whether the
predicted sample is a positive or a negative sample. Precision calculates how many of the positive
samples are actually positive, and the Recall calculates how many of the positive examples in the
sample are predicted correctly. Furthermore, F 1 is the harmonic average of them, which is the
most commonly used evaluation metrics.
3.2.2 Multi-label Metrics. Compared with single-label text classification, multi-label text classi-
fication divides the text into multiple category labels, and the number of category labels is variable.
These metrics are designed for single label text classification, which are not suitable for multi-label
tasks. Thus, there are some metrics designed for multi-label text classification.
Micro−F 1. The Micro−F 1 [283] is a measure that considers the overall accuracy and recall of all
labels. The Micro−F 1 is defined as:
2Pt × R t
Micro−F 1 = , (6)
P +R
where:  
T Pt T Pt
P =  t ∈S , R =  t ∈S . (7)
t ∈S T P t + F P t t ∈S T P t + F N t

Macro−F 1. The Macro−F 1 [283] calculates the average F 1 of all labels. Unlike Micro−F 1, which
sets even weight to every example, Macro −F 1 sets the same weight to all labels in the average
process. Formally, Macro−F 1 is defined as:
1  2Pt × R t
Macro−F 1 = , (8)
S Pt + R t
t ∈S

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:25

where:
T Pt T Pt
Pt = , Rt = . (9)
T Pt + F Pt T Pt + F Nt
In addition to the above evaluation metrics, there are some rank-based evaluation metrics for
extreme multi-label classification tasks, including P@K and N DCG@K.
Precision at Top K (P@K). The P@K [96] is the precision at the top k. For P@K, each text
has a set of L ground truth labels Lt = {l 0 , l 1 , l 2 . . . , l L−1 }, in order of decreasing probability
Pt = [p0 , p1 , p2 . . . , pQ −1 ]. The precision at k is
L,k )−1
min(
1
P@K = rel Li (Pt (j)) , (10)
k j=0

⎧
⎪ 1 if p ∈ L
relL (p) = ⎨
⎪ 0 otherwise , (11)
⎩
where L is the number of ground truth labels or possible answers on each text and k is the number
of selected labels on extreme multi-label text classification.
Normalized Discounted Cumulated Gains (NDCG@K). The N DCG@K [96] is

1 
n−1
rel Li (Pt (j))
N DCG@K = , (12)
I DCG (Li , k ) j=0 ln(j + 1)

where I DCG is ideal discounted cumulative gain and the particular rank position n is
n = min (max (|Pi | , |Li |) , k ) . (13)
Among these multi-label evaluation metrics, Micro−F 1 considers the number of categories, which
makes it suitable for the unbalanced data distribution. Macro −F 1 does not take into account the
amount of data that treats each class equally. Thus, it is easily affected by the classes with high
Recall and Precision. When the number of categories is large or extremely large, either P@K or
NDCG@K is used.

4 QUANTITATIVE RESULTS
There are many differences between sentiment analysis, news classification, topic labeling, and
natural language inference tasks, which can not be simplified modeled as a text classification task.
In this section, we tabulate the performance of the main models given in their articles on classic
datasets evaluated by classification accuracy, as shown in Table 4, including MR, SST-2, IMDB,
Yelp.P, Yelp.F, Amazon.F, 20NG, AG, DBpedia, and SNLI.
We give the performance of NB and SVM algorithms from RNTN [64] due to that the less tra-
ditional text classification model has been an experiment on datasets in Table 4. The accuracy of
NB and SVM are 81.8% and 79.4% on SST-2, respectively. We can see that, in the SST-2 data set
with only two categories, the accuracy of NB is better than that of SVM. It may be because NB
has relatively stable classification efficiency on new datasets. The performance is also stable on
small datasets. Compared with the deep learning model, the performance of NB is lower. NB has
the advantage of lower computational complexity than deep models. However, it requires manual
classification features, making it difficult to migrate the model directly to other data sets.
For deep learning models, pre-trained models get better results on most datasets. It means that
if you need to implement a text classification task, you can preferentially try pre-trained models,

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:26 Q. Li et al.

Table 4. Accuracy of Text Classification Models on Primary Datasets Evaluated by

Classification Accuracy (in Terms of Publication Year)
Sentiment News Topic NLI
Model
MR
SST-2 IMDB Yelp.P Yelp.F Amz.F 20NG AG DBpedia SNLI
NB [8] 81.80- - - - - - - - -
SVM [42] 79.40- - - - - - - - -
Tree-CRF [284] -77.30 - - - - - - - -
RAE [60] 77.70
82.40 - - - - - - - -
MV-RNN [62] 79.00
82.90 - - - - - - - -
RNTN [64] 75.90
85.40 - - - - - - - -
DCNN [5] 86.80 89.40 - - - - - - -
Paragraph-Vec [67] 87.80 92.58 - - - - - - -
TextCNN[17] 81.50 88.10 - - - - - - - -
TextRCNN [72] - - - - - - 96.49 - - -
DAN [69] - 86.30 89.40 - - - - - - -
Tree-LSTM [1] 88.00 - - - - - - - -
CharCNN [93] - - - 95.12 62.05 - - 90.49 98.45 -
HAN [108] - - 49.40 - - 63.60 - - - -
SeqTextRCNN [7] - - - - - - - - - -
oh-2LSTMp [75] - - 94.10 97.10 67.61 - 86.68 93.43 99.16 -
LSTMN [112] - 87.30 - - - - - - - -
Multi-Task [79] - 87.90 91.30 - - - - - - -
BLSTM-2DCNN [77] 82.30 89.50 - - - - 96.50 - - -
TopicRNN [83] - - 93.72 - - - - - - -
DPCNN [98] - - - 97.36 69.42 65.19 - 93.13 99.12 -
KPCNN [100] 83.25 - - - - - - 88.36 - -
RNN-Capsule [87] 83.80 - - - - - - - -
ULMFiT [285] - - 95.40 97.84 71.02 - - 94.99 99.20 -
LEAM[286] 76.95 - - 95.31 64.09 - 81.91 92.45 99.02 -
TextCapsule [101] 82.30 86.80 - - - - - 92.60 - -
TextGCN [155] 76.74 - - - - - 86.34 67.61 - -
BERT-base [19] - 93.50 95.63 98.08 70.58 61.60 - - - 91.00
BERT-large [19] - 94.90 95.79 98.19 71.38 62.20 - - - 91.70
MT-DNN[287] - 95.60 83.20 - - - - - - 91.50
XLNet-Large [138] - 96.80 96.21 98.45 72.20 67.74 - - - -
XLNet [138] - 97.00 - - - - - 95.51 99.38 -
RoBERTa [140] - 96.40 - - - - - - - 92.60
Bold is the most accurate.

such as BERT, RoBERTa, and XLNET, etc., except MR and 20NG, which have not been experi-
mented on BERT-based models. Pre-trained models are essential to NLP. It uses a deep model
to learn a better feature of the text. It also demonstrates that the accuracy of NLP tasks can be
significantly improved by a profound model that can be pre-trained from unlabeled datasets. For
the MR dataset, the accuracy of RNN-Capsule [87] is 83.8%, obtaining the best result. It suggests
that RNN-Capsule builds a capsule in each category for sentiment analysis. It can output words
including sentiment trends indicating attributes of capsules with no applying linguistic knowl-
edge. For 20NG dataset, BLSTM-2DCNN [77] gets 96.5% score with the best accuracy score. It
may demonstrate the effectiveness of applying the 2D max-pooling operation to obtain a fixed-
length representation of the text and utilize 2D convolution to sample more meaningful matrix
information.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:27

5 FUTURE RESEARCH CHALLENGES

Text classification – as efficient information retrieval and mining technology – plays a vital role in
managing text data. It uses NLP, data mining, machine learning, and other techniques to automat-
ically classify and discover different text types. Text classification takes multiple types of text as
input, and the text is represented as a vector by the pre-training model. Then the vector is fed into
the DNN for training until the termination condition is reached, and finally, the performance of
the training model is verified by the downstream task. Existing models have already shown their
usefulness in text classification, but there are still many possible improvements to explore.
Although some new text classification models repeatedly brush up the accuracy index of most
classification tasks, they cannot indicate whether the model “understands” the text from the se-
mantic level like human beings can. Moreover, with the emergence of the noise sample, the small
sample noise may cause the decision confidence to change substantially or even lead to decision
reversal. Therefore, the semantic representation ability and robustness of the model need to be
proven in practice. Besides, the pre-trained semantic representation model represented by word
vectors can often improve the performance of downstream NLP tasks. The existing research on
the transfer strategy of context-free word vectors is still relatively preliminary. Thus, we conclude
from a data, models, and performance perspective, text classification mainly faces the following
challenges.

5.1 Challenges from Data Perspective

For a text classification task, data is essential to model performance, whether it is a traditional
or deep learning method. The text data mainly studied includes multi-chapter, short text, cross-
language, multi-label, less sample text. For the characteristics of these data, the existing technical
challenges are as follows:
Zero-shot/Few-shot learning. Zero-shot or few-shot learning for text classification aims to clas-
sify text having no or few same labeled class data. However, the current models are too dependent
on numerous labeled data. The performance of these models is significantly affected by zero-shot
or few-shot learning. Thus, some works focus on tackling these problems. The main idea is to infer
the features through learning kinds of semantic knowledge, such as learning relationships among
classes [288] and incorporating class descriptions [170]. Furthermore, latent features generation
[289], meta-Learning [106, 290, 291], and dynamic memory mechanism [292] are also efficient
methods. Nevertheless, with the limitation of little unseen class data and different data distribu-
tion between seen class and unseen class, there is still a long way to go to reach the learning ability
comparable to that of humans.
The external knowledge. As we all know, the more beneficial information is input into a DNN,
the better the performance. For example, a question answering system incorporating common-
sense knowledge can answer questions about the real world and help solve problems with incom-
plete information. Therefore, adding external knowledge (knowledge base or knowledge graph)
[293, 294] is an efficient way to promote the model’s performance. The existing knowledge in-
cludes conceptual information [100, 127, 204], commonsense knowledge [223], knowledge base
information [295, 296], general knowledge graph [170], and so on, which enhances the semantic
representation of texts. Nevertheless, with the limitation of input scale, how and what to add for
different tasks is still a challenge.
Special domain with many terminologies. Most of the existing models are supervised models,
which over-rely on numerous labeled data. When the sample size is too small, or zero samples
occur, the performance of the model will be significantly affected. New data set annotation takes a
lot of time. Therefore, unsupervised learning and semi-supervised learning have great potential for

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:28 Q. Li et al.

text classification. Furthermore, texts in a particular field [297, 298], such as financial and medical
texts, contain many specific words or domain experts intelligible slang, abbreviations, etc., which
make the existing pre-trained word vectors challenging to work on.
The multi-label text classification task. Multi-label text classification requires full considera-
tion of the semantic relationship among labels, and the embedding and encoding of the model is a
process of lossy compression [299, 300]. Therefore, how to reduce the loss of hierarchical seman-
tics and retain rich and complex document semantic information during training is still a problem
to be solved.

5.2 Challenges from Model Perspective

Most existing structures of traditional and deep learning models are tried for text classification,
including integration methods. BERT learns a language representation that can be used to fine-
tune for many NLP tasks. The primary method is to increase data, improve computation power,
and design training procedures for getting better results [301–303]. Understanding the tradeoff
between data and compute resources and prediction performance is worth studying.
Text representation. The text representation method based on the vector space model is simple
and effective in the text preprocessing stage. However, it will lose the semantic information of
the text, so the application performance based on this method is limited. The proposed seman-
tically based text representation method is too time-consuming. Therefore, the efficient semanti-
cally based text representation method still needs further research. In the text representation of
text classification based on deep learning, word embedding is the main concept, while the repre-
sentation unit is described differently in different languages. Then, a word is represented in the
form of a vector by learning mapping rules through the model. Therefore, how to design adaptive
data representation methods is more conducive to the combination of deep learning and specific
classification tasks.
Model integration. Most structures of traditional and deep learning models are tried for text
classification, including integration methods. RNN requires recursive step by step to get global
information. CNN can obtain local information, and the sensing field can be increased through the
multi-layer stack to capture more comprehensive contextual information. Attention mechanisms
learn global dependency among words in a sentence. The transformer model is dependent on
attention mechanisms to establish the depth of the global dependency relationship between the
input and output. Therefore, designing an integrated model is worth trying to take advantage of
these models.
Model efficiency. Text classification models based on deep learning are highly effective, such as
CNNs, RNNs, and GNNs. However, there are many technical limitations, such as the depth of the
network layer, regularization problem, network learning rate, etc. Therefore, there is still more
broad space for development to optimize the algorithm and improve the speed of model training.

5.3 Challenges from Performance Perspective

The traditional model and the deep model can achieve good performance in most text classification
tasks, but the anti-interference ability of their results needs to be improved [176, 304, 305]. How
to realize the interpretation of the deep model is also a technical challenge.
The semantic robustness of the model. In recent years, researchers have designed many models
to enhance the accuracy of text classification models. However, when there are some adversarial
samples in the datasets, the model’s performance decreases significantly. Adversarial training is
a crucial method to improve the robustness of the pre-training model. For example, a popular

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:29

approach is converting attack into defense and using the adversarial sample training model. Con-
sequently, how to improve the robustness of models is a current research hotspot and challenge.
The interpretability of the model. DNNs have unique advantages in feature extraction and
semantic mining and have achieved excellent text classification tasks. Only a better understanding
of the theories behind these models can accurately design better models for various applications.
However, deep learning is a black-box model, the training process is challenging to reproduce,
and the implicit semantics and output interpretability are poor. It makes the improvement and
optimization of the model lose clear guidelines. Why does one model outperform another on one
data set but underperform on others? What does the deep learning model learn? Furthermore, we
cannot accurately explain why the model improves performance.

6 CONCLUSION
This paper principally introduces the existing models for text classification tasks from traditional
models to deep learning. Firstly, we introduce some primary traditional models and deep learn-
ing models with a summary table. The traditional model improves text classification performance
mainly by improving the feature extraction scheme and classifier design. In contrast, the deep
learning model enhances performance by improving the presentation learning method, model
structure, and additional data and knowledge. Then, we introduce the datasets with a summary
table and evaluation metrics for single-label and multi-label tasks. Furthermore, we give the quan-
titative results of the leading models in a summary table under different applications for classic
text classification datasets. Finally, we summarize the possible future research challenges of text
classification.

ACKNOWLEDGMENTS
Thanks for computing infrastructure provided by Huawei MindSpore platform.

REFERENCES
[1] Kai Sheng Tai, Richard Socher, and Christopher D. Manning. 2015. Improved semantic representations from tree-
structured long short-term memory networks. In Proc. ACL, 2015. 1556–1566. http://dx.doi.org/10.3115/v1/p15-1150
[2] Xiao-Dan Zhu, Parinaz Sobhani, and Hongyu Guo. 2015. Long short-term memory over recursive structures. In Proc.
ICML, 2015. 1604–1612. http://proceedings.mlr.press/v37/zhub15.html.
[3] Junyang Chen, Zhiguo Gong, and Weiwen Liu. 2020. A dirichlet process biterm-based mixture model for short text
stream clustering. Appl. Intell. 50, 5 (2020), 1609–1619. http://dx.doi.org/10.1007/s10489-019-01606-1
[4] Junyang Chen, Zhiguo Gong, and Weiwen Liu. 2019. A nonparametric model for online topic discovery with word
embeddings. Inf. Sci. 504 (2019), 32–47. http://dx.doi.org/10.1016/j.ins.2019.07.048
[5] Nal Kalchbrenner, Edward Grefenstette, and Phil Blunsom. 2014. A convolutional neural network for modelling
sentences. In Proc. ACL, 2014. 655–665. http://dx.doi.org/10.3115/v1/p14-1062
[6] Pengfei Liu, Xipeng Qiu, Xinchi Chen, Shiyu Wu, and Xuanjing Huang. 2015. Multi-timescale long short-term mem-
ory neural network for modelling sentences and documents. In Proc. EMNLP, 2015. 2326–2335. http://dx.doi.org/10.
18653/v1/d15-1280
[7] Ji Young Lee and Franck Dernoncourt. 2016. Sequential short-text classification with recurrent and convolutional
neural networks. In Proc. NAACL, 2016. 515–520. http://dx.doi.org/10.18653/v1/n16-1062
[8] M. E. Maron. 1961. Automatic indexing: An experimental inquiry. J. ACM 8, 3 (1961), 404–417. http://dx.doi.org/10.
1145/321075.321084
[9] Thomas M. Cover and Peter E. Hart. 1967. Nearest neighbor pattern classification. IEEE Trans. Inf. Theory 13, 1 (1967),
21–27. http://dx.doi.org/10.1109/TIT.1967.1053964
[10] Thorsten Joachims. 1998. Text categorization with support vector machines: Learning with many relevant features.
In Proc. ECML, 1998. 137–142. http://dx.doi.org/10.1007/BFb0026683
[11] Rami Aly, Steffen Remus, and Chris Biemann. 2019. Hierarchical multi-label classification of text with capsule net-
works. In Proceedings of the 57th Conference of the Association for Computational Linguistics, ACL 2019, Florence, Italy,
July 28 - August 2, 2019, Volume 2: Student Research Workshop. 323–330. http://dx.doi.org/10.18653/v1/p19-2045

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:30 Q. Li et al.

[12] Kamran Kowsari, Kiana Jafari Meimandi, Mojtaba Heidarysafa, Sanjana Mendu, Laura E. Barnes, and Donald E.
Brown. 2019. Text classification algorithms: A survey. Information 10, 4 (2019), 150. http://dx.doi.org/10.3390/
info10040150
[13] Shervin Minaee, Nal Kalchbrenner, Erik Cambria, Narjes Nikzad, Meysam Chenaghlu, and Jianfeng Gao. 2020. Deep
learning based text classification: A comprehensive review. CoRR abs/2004.03705 (2020). arXiv:2004.03705 https://
arxiv.org/abs/2004.03705.
[14] Leo Breiman. 2001. Random forests. Mach. Learn. 45, 1 (2001), 5–32. http://dx.doi.org/10.1023/A:1010933404324
[15] Tianqi Chen and Carlos Guestrin. 2016. XGBoost: A scalable tree boosting system. In Proc. ACM SIGKDD, 2016. 785–
794. http://dx.doi.org/10.1145/2939672.2939785
[16] Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. Light-
GBM: A highly efficient gradient boosting decision tree. In Proc. NeurIPS, 2017. 3146–3154. http://papers.nips.cc/
paper/6907-lightgbm-a-highly-efficient-gradient-boosting-decision-tree.
[17] Yoon Kim. 2014. Convolutional neural networks for sentence classification. In Proc. EMNLP, 2014. 1746–1751. http:
//dx.doi.org/10.3115/v1/d14-1181
[18] Saad Albawi, Tareq Abed Mohammed, and Saad Al-Zawi. 2017. Understanding of a convolutional neural network.
In 2017 International Conference on Engineering and Technology (ICET). IEEE, 1–6.
[19] Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional
transformers for language understanding. In Proc. NAACL, 2019. 4171–4186. https://www.aclweb.org/anthology/N19-
1423/.
[20] Yin Zhang, Rong Jin, and Zhi-Hua Zhou. 2010. Understanding bag-of-words model: A statistical framework. Inter-
national Journal of Machine Learning and Cybernetics 1, 1–4 (2010), 43–52.
[21] William B. Cavnar, John M. Trenkle, et al. 1994. N-gram-based text categorization. In Proceedings of SDAIR-94, 3rd
Annual Symposium on Document Analysis and Information Retrieval, Vol. 161175. Citeseer.
[22] Ricardo Baeza-Yates, and Berthier Ribeiro-Neto. 1999. Modern information retrieval. ACM press, Vol. 463.
[23] Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in
vector space. In Proc. ICLR, 2013. http://arxiv.org/abs/1301.3781.
[24] Jeffrey Pennington, Richard Socher, and Christopher D. Manning. 2014. GloVe: Global vectors for word representa-
tion. In Proc. EMNLP, 2014. 1532–1543. http://dx.doi.org/10.3115/v1/d14-1162
[25] Min-Ling Zhang and Kun Zhang. 2010. Multi-label learning by exploiting label dependency. In Proc. ACM SIGKDD,
2010. 999–1008. http://dx.doi.org/10.1145/1835804.1835930
[26] Karl-Michael Schneider. 2004. A new feature selection score for multinomial Naïve Bayes text classification based
on KL-divergence. In Proc. ACL, 2004. https://www.aclweb.org/anthology/P04-3024/.
[27] Thomas M. Cover and Joy A. Thomas. 2006. Elements of Information Theory (Wiley Series in Telecommunications and
Signal Processing). Wiley-Interscience, USA.
[28] Wenyuan Dai, Gui-Rong Xue, Qiang Yang, and Yong Yu. 2007. Transferring Naïve Bayes classifiers for text classifi-
cation. In Proc. AAAI, 2007. 540–545. http://www.aaai.org/Library/AAAI/2007/aaai07-085.php.
[29] A. P. Dempster, N. M. Laird, and D. B. Rubin. 1977. Maximum likelihood from incomplete data via the EM algorithm.
Journal of the Royal Statistical Society (1977).
[30] M. Granik and V. Mesyura. 2017. Fake news detection using Naïve Bayes classifier. In 2017 IEEE First Ukraine Confer-
ence on Electrical and Computer Engineering (UKRCON). 900–903. http://dx.doi.org/10.1109/UKRCON.2017.8100379
[31] Mohamad Syahrul Mubarok, Kang Adiwijaya, and Muhammad Aldhi. 2017. Aspect-based sentiment analysis to re-
view products using Naïve Bayes. AIP Conference Proceedings 1867 (08 2017), 020060. http://dx.doi.org/10.1063/1.
4994463
[32] Shuo Xu. 2018. Bayesian Naïve Bayes classifiers to text classification. J. Inf. Sci. 44, 1 (2018), 48–59. http://dx.doi.org/
10.1177/0165551516677946
[33] G. Singh, B. Kumar, L. Gaur, and A. Tyagi. 2019. Comparison between multinomial and Bernoulli Naïve Bayes for text
classification. In 2019 International Conference on Automation, Computational and Technology Management (ICACTM).
593–596. http://dx.doi.org/10.1109/ICACTM.2019.8776800
[34] 2007. 20NG Corpus. http://ana.cachopo.org/datasets-for-single-label-text-categorization. (2007).
[35] Mark Craven, Dan DiPasquo, Dayne Freitag, Andrew McCallum, Tom M. Mitchell, Kamal Nigam, and Seán Slattery.
1998. Learning to extract symbolic knowledge from the World Wide Web. In Proceedings of the Fifteenth National
Conference on Artificial Intelligence and Tenth Innovative Applications of Artificial Intelligence Conference, AAAI 98,
IAAI 98, July 26-30, 1998, Madison, Wisconsin, USA. 509–516. http://www.aaai.org/Library/AAAI/1998/aaai98-072.
php.
[36] T. Jo. 2017. Using K nearest neighbors for text segmentation with feature similarity. In 2017 International Confer-
ence on Communication, Control, Computing and Electronics Engineering (ICCCCEE). 1–5. http://dx.doi.org/10.1109/
ICCCCEE.2017.7866706

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:31

[37] Li Baoli, Lu Qin, and Yu Shiwen. 2004. An adaptive <i>k</i>-nearest neighbor text categorization strategy. ACM
Transactions on Asian Language Information Processing 3, 4 (Dec. 2004), 215–226. http://dx.doi.org/10.1145/1039621.
1039623
[38] Shufeng Chen. 2018. K-nearest neighbor algorithm optimization in text categorization. IOP Conference Series: Earth
and Environmental Science 108 (Jan 2018), 052074. http://dx.doi.org/10.1088/1755-1315/108/5/052074
[39] Shengyi Jiang, Guansong Pang, Meiling Wu, and Limin Kuang. 2012. An improved K-nearest-neighbor algorithm for
text categorization. Expert Syst. Appl. 39, 1 (2012), 1503–1509. http://dx.doi.org/10.1016/j.eswa.2011.08.040
[40] Pascal Soucy and Guy W. Mineau. 2001. A simple KNN algorithm for text categorization. In Proc. ICDM, 2001. 647–648.
http://dx.doi.org/10.1109/ICDM.2001.989592
[41] Songbo Tan. 2005. Neighbor-weighted K-nearest neighbor for unbalanced text corpus. Expert Syst. Appl. 28, 4 (2005),
667–671. http://dx.doi.org/10.1016/j.eswa.2004.12.023.
[42] Corinna Cortes and Vladimir Vapnik. 1995. Support-vector networks. Mach. Learn. 20, 3 (1995), 273–297. http://dx.
doi.org/10.1007/BF00994018
[43] Christina Leslie, Eleazar Eskin, and William Stafford Noble. 2001. The spectrum kernel: A string kernel for SVM
protein classification. In Biocomputing 2002. World Scientific, 564–575.
[44] Hirotoshi Taira and Masahiko Haruno. 1999. Feature selection in SVM text categorization. In AAAI/IAAI. 480–486.
[45] Xin Li and Yuhong Guo. 2013. Active learning with multi-label SVM classification. In IJCAI. Citeseer, 1479–1485.
[46] Tao Peng, Wanli Zuo, and Fengling He. 2008. SVM based adaptive learning method for text classification from positive
and unlabeled documents. Knowledge and Information Systems 16, 3 (2008), 281–301.
[47] Thorsten Joachims. 2001. A statistical learning model of text classification for support vector machines. In Proc. SIGIR,
2001. 128–136. http://dx.doi.org/10.1145/383952.383974
[48] T. Joachims. 1999. Transductive inference for text classification using support vector machines. In International Con-
ference on Machine Learning.
[49] Tom M. Mitchell. 1997. Machine Learning. McGraw-Hill. http://www.worldcat.org/oclc/61321007.
[50] Rajeev Rastogi and Kyuseok Shim. 2000. PUBLIC: A decision tree classifier that integrates building and pruning.
Data Min. Knowl. Discov. 4, 4 (2000), 315–344. http://dx.doi.org/10.1023/A:1009887311454
[51] Ross J. Quinlan. 1986. Induction of decision trees. Machine Learning 1, 1 (1986), 81–106.
[52] J. Ross Quinlan. 1993. C4.5: Programs for Machine Learning. Morgan Kaufmann Publishers Inc., San Francisco, CA,
USA.
[53] Micheline Kamber, Lara Winstone, Wan Gong, Shan Cheng, and Jiawei Han. 1997. Generalization and decision tree
induction: Efficient classification in data mining. In Proceedings Seventh International Workshop on Research Issues in
Data Engineering. High Performance Database Management for Large-Scale Applications. IEEE, 111–120.
[54] David E. Johnson, Frank J. Oles, Tong Zhang, and Thilo Götz. 2002. A decision-tree-based symbolic rule induction
system for text categorization. IBM Syst. J. 41, 3 (2002), 428–437. http://dx.doi.org/10.1147/sj.413.0428
[55] Peerapon Vateekul and Miroslav Kubat. 2009. Fast induction of multiple decision trees in text categorization from
large scale, imbalanced, and multi-label data. In Proc. ICDM Workshops, 2009. 320–325. http://dx.doi.org/10.1109/
ICDMW.2009.94
[56] Yoav Freund and Robert E. Schapire. 1995. A decision-theoretic generalization of on-line learning and an application
to boosting. In Proc. EuroCOLT, 1995. 23–37. http://dx.doi.org/10.1007/3-540-59119-2_166
[57] Robert E. Schapire and Yoram Singer. 1999. Improved boosting algorithms using confidence-rated predictions. Mach.
Learn. 37, 3 (1999), 297–336. http://dx.doi.org/10.1023/A:1007614523901
[58] Ameni Bouaziz, Christel Dartigues-Pallez, Célia da Costa Pereira, Frédéric Precioso, and Patrick Lloret. 2014. Short
text classification using semantic random forest. In Proc. DAWAK, 2014. 288–299. http://dx.doi.org/10.1007/978-3-319-
10160-6_26
[59] Md. Zahidul Islam, Jixue Liu, Jiuyong Li, Lin Liu, and Wei Kang. 2019. A semantics aware random forest for text
classification. In Proc. CIKM, 2019. 1061–1070. http://dx.doi.org/10.1145/3357384.3357891
[60] Richard Socher, Jeffrey Pennington, Eric H. Huang, Andrew Y. Ng, and Christopher D. Manning. 2011. Semi-
supervised recursive autoencoders for predicting sentiment distributions. In Proc. EMNLP, 2011. 151–161. https:
//www.aclweb.org/anthology/D11-1014/.
[61] 2011. A MATLAB Implementation of RAE. https://github.com/vin00/Semi-Supervised-Recursive-Autoencoders-for-
Predicting-Sentiment-Distributions. (2011).
[62] Richard Socher, Brody Huval, Christopher D. Manning, and Andrew Y. Ng. 2012. Semantic compositionality through
recursive matrix-vector spaces. In Proc. EMNLP, 2012. 1201–1211. https://www.aclweb.org/anthology/D12-1110/.
[63] 2012. A Tensorflow Implementation of MV_RNN. https://github.com/github-pengge/MV_RNN. (2012).
[64] Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Christopher D. Manning, Andrew Y. Ng, and Christopher
Potts. 2013. Recursive deep models for semantic compositionality over a sentiment treebank. In Proc. EMNLP, 2013.
1631–1642. https://www.aclweb.org/anthology/D13-1170/.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:32 Q. Li et al.

[65] 2013. A MATLAB Implementation of RNTN. https://github.com/pondruska/DeepSentiment. (2013).

[66] Ozan Irsoy and Claire Cardie. 2014. Deep recursive neural networks for compositionality in language. In Proc.
NIPS, 2014. 2096–2104. http://papers.nips.cc/paper/5551-deep-recursive-neural-networks-for-compositionality-in-
language.
[67] Quoc V. Le and Tomas Mikolov. 2014. Distributed representations of sentences and documents. In Proc. ICML, 2014.
1188–1196. http://proceedings.mlr.press/v32/le14.html.
[68] 2014. A PyTorch Implementation of Paragraph Vectors (doc2vec). https://github.com/inejc/paragraph-vectors. (2014).
[69] Mohit Iyyer, Varun Manjunatha, Jordan L. Boyd-Graber, and Hal Daumé III. 2015. Deep unordered composition rivals
syntactic methods for text classification. In Proc. ACL, 2015. 1681–1691. http://dx.doi.org/10.3115/v1/p15-1162
[70] 2015. An Implementation of DAN. https://github.com/miyyer/dan. (2015).
[71] 2015. A PyTorch Implementation of Tree-LSTM. https://github.com/stanfordnlp/treelstm. (2015).
[72] Siwei Lai, Liheng Xu, Kang Liu, and Jun Zhao. 2015. Recurrent convolutional neural networks for text classification
(AAAI’15). AAAI Press, 2267–2273.
[73] 2015. A Tensorflow Implementation of TextRCNN. https://github.com/roomylee/rcnn-text-classification. (2015).
[74] 2015. An Implementation of MT-LSTM. https://github.com/AlexAntn/MTLSTM. (2015).
[75] Rie Johnson and Tong Zhang. 2016. Supervised and semi-supervised text categorization using LSTM for region
embeddings. In Proc. ICML, 2016. 526–534. http://proceedings.mlr.press/v48/johnson16.html.
[76] 2015. An Implementation of oh-2LSTMp. http://riejohnson.com/cnn_20download.html. (2015).
[77] Peng Zhou, Zhenyu Qi, Suncong Zheng, Jiaming Xu, Hongyun Bao, and Bo Xu. 2016. Text classification improved
by integrating bidirectional LSTM with two-dimensional max pooling. In Proc. COLING, 2016. 3485–3495. https:
//www.aclweb.org/anthology/C16-1329/.
[78] 2016. An Implementation of BLSTM-2DCNN. https://github.com/ManuelVs/NNForTextClassification. (2016).
[79] Pengfei Liu, Xipeng Qiu, and Xuanjing Huang. 2016. Recurrent neural network for text classification with multi-task
learning. In Proc. IJCAI, 2016. 2873–2879. http://www.ijcai.org/Abstract/16/408.
[80] 2016. A PyTorch Implementation of Multi-Task. https://github.com/baixl/text_classification. (2016).
[81] Bjarke Felbo, Alan Mislove, Anders Søgaard, Iyad Rahwan, and Sune Lehmann. 2017. Using millions of emoji oc-
currences to learn any-domain representations for detecting sentiment, emotion and sarcasm. In Proc. EMNLP, 2017.
1615–1625. https://www.aclweb.org/anthology/D17-1169/.
[82] 2018. A Keras Implementation of DeepMoji. https://github.com/bfelbo/DeepMoji. (2018).
[83] Adji B. Dieng, Chong Wang, Jianfeng Gao, and John W. Paisley. 2017. TopicRNN: A recurrent neural network with
long-range semantic dependency. In Proc. ICLR, 2017. https://openreview.net/forum?id=rJbbOLcex.
[84] 2017. A PyTorch Implementation of TopicRNN. https://github.com/dangitstam/topic-rnn. (2017).
[85] Takeru Miyato, Andrew M. Dai, and Ian J. Goodfellow. 2017. Adversarial training methods for semi-supervised text
classification. In Proc. ICLR, 2017. https://openreview.net/forum?id=r1X3g2_xl.
[86] 2017. A Tensorflow Implementation of Virtual Adversarial Training. https://github.com/tensorflow/models/tree/
master/adversarial_text. (2017).
[87] Yequan Wang, Aixin Sun, Jialong Han, Ying Liu, and Xiaoyan Zhu. 2018. Sentiment analysis by capsules. In Proc.
WWW, 2018. 1165–1174. http://dx.doi.org/10.1145/3178876.3186015
[88] 2018. A PyTorch Implementation of RNN-Capsule. https://github.com/wangjiosw/Sentiment-Analysis-by-Capsules.
(2018).
[89] Yi Zhao, Yanyan Shen, and Junjie Yao. 2019. Recurrent neural network for text classification with hierarchical mul-
tiscale dense connections. In Proceedings of the Twenty-Eighth International Joint Conference on Artificial Intelligence,
IJCAI 2019, Macao, China, August 10-16, 2019. 5450–5456. http://dx.doi.org/10.24963/ijcai.2019/757
[90] 2019. An Implementation of HM-DenseRNNs. https://github.com/zhaoyizhaoyi/hm-densernns. (2019).
[91] 2014. A Keras Implementation of TextCNN. https://github.com/alexander-rakhlin/CNN-for-Sentence-Classification-
in-Keras. (2014).
[92] 2014. A Tensorflow Implementation of DCNN. https://github.com/kinimod23/ATS_Project. (2014).
[93] Xiang Zhang, Junbo Jake Zhao, and Yann LeCun. 2015. Character-level convolutional networks for text classification.
In Proc. NeurIPS, 2015. 649–657. http://papers.nips.cc/paper/5782-character-level-convolutional-networks-for-text-
classification.
[94] 2015. A Tensorflow Implementation of CharCNN. https://github.com/mhjabreel/CharCNN. (2015).
[95] 2016. A Keras Implementation of SeqTextRCNN. https://github.com/ilimugur/short-text-classification. (2016).
[96] Jingzhou Liu, Wei-Cheng Chang, Yuexin Wu, and Yiming Yang. 2017. Deep learning for extreme multi-label text
classification. In Proc. ACM SIGIR, 2017. 115–124. http://dx.doi.org/10.1145/3077136.3080834
[97] 2017. A Pytorch Implementation of XML-CNN. https://github.com/siddsax/XML-CNN. (2017).
[98] Rie Johnson and Tong Zhang. 2017. Deep pyramid convolutional neural networks for text categorization. In Proc.
ACL, 2017. 562–570. http://dx.doi.org/10.18653/v1/P17-1052

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:33

[99] 2017. A PyTorch Implementation of DPCNN. https://github.com/Cheneng/DPCNN. (2017).

[100] Jin Wang, Zhongyuan Wang, Dawei Zhang, and Jun Yan. 2017. Combining knowledge with deep convolutional neural
networks for short text classification. In Proc. IJCAI, 2017. 2915–2921. http://dx.doi.org/10.24963/ijcai.2017/406
[101] Min Yang, Wei Zhao, Jianbo Ye, Zeyang Lei, Zhou Zhao, and Soufei Zhang. 2018. Investigating capsule networks
with dynamic routing for text classification. In Proceedings of the 2018 Conference on Empirical Methods in Natural
Language Processing, Brussels, Belgium, October 31 - November 4, 2018. 3110–3119. http://dx.doi.org/10.18653/v1/d18-
1350
[102] 2018. A Tensorflow Implementation of TextCapsule. https://github.com/andyweizhao/capsule_text_classification.
(2018).
[103] Kazuya Shimura, Jiyi Li, and Fumiyo Fukumoto. 2018. HFT-CNN: Learning hierarchical category structure for multi-
label short text categorization. In Proc. EMNLP, 2018. 811–816. http://dx.doi.org/10.18653/v1/d18-1093
[104] 2018. An Implementation of HFT-CNN. https://github.com/ShimShim46/HFT-CNN. (2018).
[105] Jingyun Xu and Yi Cai. 2019. Incorporating context-relevant knowledge into convolutional neural networks for
short text classification. In The Thirty-Third AAAI Conference on Artificial Intelligence, AAAI 2019, The Thirty-First
Innovative Applications of Artificial Intelligence Conference, IAAI 2019, The Ninth AAAI Symposium on Educational
Advances in Artificial Intelligence, EAAI 2019, Honolulu, Hawaii, USA, January 27 - February 1, 2019. 10067–10068.
http://dx.doi.org/10.1609/aaai.v33i01.330110067
[106] Yujia Bao, Menghua Wu, Shiyu Chang, and Regina Barzilay. 2020. Few-shot text classification with distributional
signatures. In Proc. ICLR, 2020. https://openreview.net/forum?id=H1emfT4twB.
[107] 2020. A PyTorch Implementation of Few-shot Text Classification with Distributional Signatures. https://github.com/
YujiaBao/Distributional-Signatures. (2020).
[108] Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alexander J. Smola, and Eduard H. Hovy. 2016. Hierarchical atten-
tion networks for document classification. In Proc. NAACL, 2016. 1480–1489. https://www.aclweb.org/anthology/N16-
1174/.
[109] 2014. A Keras Implementation of TextCNN. https://github.com/richliao/textClassifier. (2014).
[110] Xinjie Zhou, Xiaojun Wan, and Jianguo Xiao. 2016. Attention-based LSTM network for cross-lingual sentiment clas-
sification. In Proc. EMNLP, 2016. 247–256. http://dx.doi.org/10.18653/v1/d16-1024
[111] 2013. NLP&CC Corpus. http://tcci.ccf.org.cn/conference/2013/index.html. (2013).
[112] Jianpeng Cheng, Li Dong, and Mirella Lapata. 2016. Long short-term memory-networks for machine reading. In Proc.
EMNLP, 2016. 551–561. http://dx.doi.org/10.18653/v1/d16-1053
[113] 2016. A Tensorflow Implementation of LSTMN. https://github.com/JRC1995/Abstractive-Summarization. (2016).
[114] Zhouhan Lin, Minwei Feng, Cícero Nogueira dos Santos, Mo Yu, Bing Xiang, Bowen Zhou, and Yoshua Bengio. 2017.
A structured self-attentive sentence embedding. In Proc. ICLR, 2017. https://openreview.net/forum?id=BJC_jUqxe.
[115] 2017. A PyTorch Implementation of Structured-Self-Attention. https://github.com/kaushalshetty/Structured-Self-
Attention. (2017).
[116] Pengcheng Yang, Xu Sun, Wei Li, Shuming Ma, Wei Wu, and Houfeng Wang. 2018. SGM: Sequence generation model
for multi-label classification. In Proc. COLING, 2018. 3915–3926. https://www.aclweb.org/anthology/C18-1330/.
[117] 2018. A PyTorch Implementation of SGM. https://github.com/lancopku/SGM. (2018).
[118] Matthew E. Peters, Mark Neumann, Mohit Iyyer, Matt Gardner, Christopher Clark, Kenton Lee, and Luke Zettle-
moyer. 2018. Deep contextualized word representations. In Proc. NAACL, 2018. 2227–2237. https://www.aclweb.org/
anthology/N18-1202/.
[119] 2018. A PyTorch Implementation of ELMo. https://github.com/flairNLP/flair. (2018).
[120] Tao Shen, Tianyi Zhou, Guodong Long, Jing Jiang, and Chengqi Zhang. 2018. Bi-directional block self-attention for
fast and memory-efficient sequence modeling. In Proc. ICLR, 2018. https://openreview.net/forum?id=H1cWzoxA-.
[121] 2018. A PyTorch Implementation of BiBloSA. https://github.com/galsang/BiBloSA-pytorch. (2018).
[122] Ronghui You, Zihan Zhang, Ziye Wang, Suyang Dai, Hiroshi Mamitsuka, and Shanfeng Zhu. 2019. AttentionXML:
Label tree-based attention-aware deep model for high-performance extreme multi-label text classification. In
Proc. NeurIPS, 2019. 5812–5822. http://papers.nips.cc/paper/8817-attentionxml-label-tree-based-attention-aware-
deep-model-for-high-performance-extreme-multi-label-text-classification.
[123] 2019. A PyTorch Implementation of AttentionXML. https://github.com/yourh/AttentionXML. (2019).
[124] Shengli Sun, Qingfeng Sun, Kevin Zhou, and Tengchao Lv. 2019. Hierarchical attention prototypical networks for
few-shot text classification. In Proc. EMNLP, 2019. 476–485. http://dx.doi.org/10.18653/v1/D19-1045
[125] Tianyu Gao, Xu Han, Zhiyuan Liu, and Maosong Sun. 2019. Hybrid attention-based prototypical networks for noisy
few-shot relation classification. In Proc. AAAI, 2019. 6407–6414. http://dx.doi.org/10.1609/aaai.v33i01.33016407
[126] 2019. A PyTorch Implementation of HATT-Proto. https://github.com/thunlp/HATT-Proto. (2019).
[127] Jindong Chen, Yizhou Hu, Jingping Liu, Yanghua Xiao, and Haiyun Jiang. 2019. Deep short text classification with
knowledge powered attention. In Proc. AAAI, 2019. 6252–6259. http://dx.doi.org/10.1609/aaai.v33i01.33016252

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:34 Q. Li et al.

[128] 2019. A PyTorch Implementation of STCKA. https://github.com/AIRobotZhang/STCKA. (2019).

[129] Kaize Ding, Jianling Wang, Jundong Li, Dingcheng Li, and Huan Liu. 2020. Be more with less: Hypergraph attention
networks for inductive text classification. In Proceedings of the 2020 Conference on Empirical Methods in Natural Lan-
guage Processing, EMNLP 2020, Online, November 16-20, 2020. 4927–4936. http://dx.doi.org/10.18653/v1/2020.emnlp-
main.399
[130] 2020. A Pytorch Implementation of HyperGAT. https://github.com/kaize0409/HyperGAT. (2020).
[131] Qipeng Guo, Xipeng Qiu, Pengfei Liu, Xiangyang Xue, and Zheng Zhang. 2020. Multi-scale self-attention for text
classification. In The Thirty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2020, The Thirty-Second Innovative
Applications of Artificial Intelligence Conference, IAAI 2020, The Tenth AAAI Symposium on Educational Advances in
Artificial Intelligence, EAAI 2020, New York, NY, USA, February 7-12, 2020. 7847–7854. https://aaai.org/ojs/index.php/
AAAI/article/view/6290.
[132] Seungtaek Choi, Haeju Park, Jinyoung Yeo, and Seung-won Hwang. 2020. Less is more: Attention supervision with
counterfactuals for text classification. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language
Processing, EMNLP 2020, Online, November 16-20, 2020. 6695–6704. http://dx.doi.org/10.18653/v1/2020.emnlp-main.
543
[133] 2019. A Tensorflow Implementation of BERT. https://github.com/google-research/bert. (2019).
[134] Ilias Chalkidis, Manos Fergadiotis, Prodromos Malakasiotis, and Ion Androutsopoulos. 2019. Large-scale multi-label
text classification on EU legislation. In Proc. ACL, 2019. 6314–6322. http://dx.doi.org/10.18653/v1/p19-1636
[135] 2019. A Tensorflow Implementation of BERT-BASE. https://github.com/iliaschalkidis/lmtc-eurlex57k. (2019).
[136] Chi Sun, Xipeng Qiu, Yige Xu, and Xuanjing Huang. 2019. How to fine-tune BERT for text classification? In Proc.
CCL, 2019. 194–206. http://dx.doi.org/10.1007/978-3-030-32381-3_16
[137] 2019. A Tensorflow Implementation of BERT4doc-Classification. https://github.com/xuyige/BERT4doc-
Classification. (2019).
[138] Zhilin Yang, Zihang Dai, Yiming Yang, Jaime G. Carbonell, Ruslan Salakhutdinov, and Quoc V. Le. 2019. XLNet:
Generalized autoregressive pretraining for language understanding. In Proc. NeurIPS, 2019. 5754–5764. http://papers.
nips.cc/paper/8812-xlnet-generalized-autoregressive-pretraining-for-language-understanding.
[139] 2019. A Tensorflow Implementation of XLNet. https://github.com/zihangdai/xlnet. (2019).
[140] Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettle-
moyer, and Veselin Stoyanov. 2019. RoBERTa: A robustly optimized BERT pretraining approach. CoRR abs/1907.11692
(2019). arXiv:1907.11692 http://arxiv.org/abs/1907.11692.
[141] 2019. A PyTorch Implementation of RoBERTa. https://github.com/pytorch/fairseq. (2019).
[142] Danilo Croce, Giuseppe Castellucci, and Roberto Basili. 2020. GAN-BERT: Generative adversarial learning for robust
text classification with a bunch of labeled examples. In Proceedings of the 58th Annual Meeting of the Association
for Computational Linguistics, ACL 2020, Online, July 5-10, 2020. 2114–2119. http://dx.doi.org/10.18653/v1/2020.acl-
main.191
[143] 2020. A Pytorch Implementation of GAN-BERT. https://github.com/crux82/ganbert. (2020).
[144] Siddhant Garg and Goutham Ramakrishnan. 2020. BAE: BERT-based adversarial examples for text classification.
In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, EMNLP 2020, Online,
November 16-20, 2020. 6174–6181. http://dx.doi.org/10.18653/v1/2020.emnlp-main.498
[145] 2020. An Implementation of BAE. https://github.com/QData/TextAttack/blob/master/textattack/attack_recipes/bae_
garg_2019.py. (2020).
[146] Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, and Radu Soricut. 2020. ALBERT:
A lite BERT for self-supervised learning of language representations. In Proc. ICLR, 2020. https://openreview.net/
forum?id=H1eA7AEtvS.
[147] 2020. A Tensorflow Implementation of ALBERT. https://github.com/google-research/ALBERT. (2020).
[148] Haopeng Zhang and Jiawei Zhang. 2020. Text graph transformer for document classification. In Proceedings of the
2020 Conference on Empirical Methods in Natural Language Processing, EMNLP 2020, Online, November 16-20, 2020.
8322–8327. http://dx.doi.org/10.18653/v1/2020.emnlp-main.668
[149] Wei-Cheng Chang, Hsiang-Fu Yu, Kai Zhong, Yiming Yang, and Inderjit S. Dhillon. 2020. Taming pretrained trans-
formers for extreme multi-label text classification. In KDD ’20: The 26th ACM SIGKDD Conference on Knowledge
Discovery and Data Mining, Virtual Event, CA, USA, August 23-27, 2020. 3163–3171. http://dx.doi.org/10.1145/3394486.
3403368
[150] 2020. An Implementation of X-Transformer. https://github.com/OctoberChang/X-Transformer. (2020).
[151] Ting Jiang, Deqing Wang, Leilei Sun, Huayi Yang, Zhengyang Zhao, and Fuzhen Zhuang. 2021. LightXML:
Transformer with dynamic negative sampling for high-performance extreme multi-label text classification. CoRR
abs/2101.03305 (2021). arXiv:2101.03305 https://arxiv.org/abs/2101.03305.
[152] 2021. An Implementation of LightXML. https://github.com/kongds/LightXML. (2021).

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:35

[153] Hao Peng, Jianxin Li, Yu He, Yaopeng Liu, Mengjiao Bao, Lihong Wang, Yangqiu Song, and Qiang Yang. 2018. Large-
scale hierarchical text classification with recursively regularized deep graph-CNN. In Proc. WWW, 2018. 1063–1072.
http://dx.doi.org/10.1145/3178876.3186005
[154] 2018. A Tensorflow Implementation of DeepGraphCNNforTexts. https://github.com/HKUST-KnowComp/
DeepGraphCNNforTexts. (2018).
[155] Liang Yao, Chengsheng Mao, and Yuan Luo. 2019. Graph convolutional networks for text classification. In Proc. AAAI,
2019. 7370–7377. http://dx.doi.org/10.1609/aaai.v33i01.33017370
[156] 2019. A Tensorflow Implementation of TextGCN. https://github.com/yao8839836/text_gcn. (2019).
[157] Felix Wu, Amauri H. Souza Jr., Tianyi Zhang, Christopher Fifty, Tao Yu, and Kilian Q. Weinberger. 2019. Simplifying
graph convolutional networks. In Proc. ICML, 2019. 6861–6871. http://proceedings.mlr.press/v97/wu19e.html.
[158] 2019. An Implementation of SGC. https://github.com/Tiiiger/SGC. (2019).
[159] Lianzhe Huang, Dehong Ma, Sujian Li, Xiaodong Zhang, and Houfeng Wang. 2019. Text level graph neural network
for text classification. In Proc. EMNLP, 2019. 3442–3448. http://dx.doi.org/10.18653/v1/D19-1345
[160] 2019. An Implementation of TextLevelGNN. https://github.com/LindgeW/TextLevelGNN. (2019).
[161] Hao Peng, Jianxin Li, Senzhang Wang, Lihong Wang, Qiran Gong, Renyu Yang, Bo Li, Philip Yu, and Lifang He. 2019.
Hierarchical taxonomy-aware and attentional graph capsule RCNNs for large-scale multi-label text classification.
IEEE Transactions on Knowledge and Data Engineering (2019).
[162] Yufeng Zhang, Xueli Yu, Zeyu Cui, Shu Wu, Zhongzhen Wen, and Liang Wang. 2020. Every document owns its
structure: Inductive text classification via graph neural networks. In Proc. ACL, 2020. 334–339. https://www.aclweb.
org/anthology/2020.acl-main.31/.
[163] 2019. A Tensorflow Implementation of TextING. https://github.com/CRIPAC-DIG/TextING. (2019).
[164] Xien Liu, Xinxin You, Xiao Zhang, Ji Wu, and Ping Lv. 2020. Tensor graph convolutional networks for text clas-
sification. In The Thirty-Fourth AAAI Conference on Artificial Intelligence, AAAI 2020, The Thirty-Second Innovative
Applications of Artificial Intelligence Conference, IAAI 2020, The Tenth AAAI Symposium on Educational Advances in
Artificial Intelligence, EAAI 2020, New York, NY, USA, February 7-12, 2020. 8409–8416. https://aaai.org/ojs/index.php/
AAAI/article/view/6359.
[165] 2019. A Tensorflow Implementation of TensorGCN. https://github.com/THUMLP/TensorGCN. (2019).
[166] Ankit Pal, Muru Selvakumar, and Malaikannan Sankarasubbu. 2020. MAGNET: Multi-label text classification using
attention-based graph neural network. In Proc. ICAART, 2020. 494–505. http://dx.doi.org/10.5220/0008940304940505
[167] 2020. A Repository of MAGNET. https://github.com/monk1337/MAGnet. (2020).
[168] 2017. A Tensorflow Implementation of Miyato et al. https://github.com/TobiasLee/Text-Classification. (2017).
[169] Jichuan Zeng, Jing Li, Yan Song, Cuiyun Gao, Michael R. Lyu, and Irwin King. 2018. Topic memory networks for
short text classification. In Proc. EMNLP, 2018. 3120–3131. http://dx.doi.org/10.18653/v1/d18-1351
[170] Jingqing Zhang, Piyawat Lertvittayakumjorn, and Yike Guo. 2019. Integrating semantic knowledge to tackle zero-
shot text classification. In Proc. NAACL, 2019. 1031–1040. http://dx.doi.org/10.18653/v1/n19-1108
[171] 2019. A Tensorflow Implementation of KG4ZeroShotText. https://github.com/JingqingZ/KG4ZeroShotText. (2019).
[172] M. K. Alsmadi, K. B. Omar, S. A. Noah, and I. Almarashdah. 2009. Performance comparison of multi-layer perceptron
(back propagation, delta rule and perceptron) algorithms in neural networks. In 2009 IEEE International Advance
Computing Conference. 296–299.
[173] Samira Pouyanfar, Saad Sadiq, Yilin Yan, Haiman Tian, Yudong Tao, Maria E. Presa Reyes, Mei-Ling Shyu, Shu-Ching
Chen, and S. S. Iyengar. 2019. A survey on deep learning: Algorithms, techniques, and applications. ACM Comput.
Surv. 51, 5 (2019), 92:1–92:36. http://dx.doi.org/10.1145/3234150
[174] Libo Qin, Wanxiang Che, Yangming Li, Minheng Ni, and Ting Liu. 2020. DCR-Net: A deep co-interactive relation net-
work for joint dialog act recognition and sentiment classification. In The Thirty-Fourth AAAI Conference on Artificial
Intelligence, AAAI 2020, The Thirty-Second Innovative Applications of Artificial Intelligence Conference, IAAI 2020, The
Tenth AAAI Symposium on Educational Advances in Artificial Intelligence, EAAI 2020, New York, NY, USA, February
7-12, 2020. 8665–8672. https://aaai.org/ojs/index.php/AAAI/article/view/6391.
[175] Zhongfen Deng, Hao Peng, Dongxiao He, Jianxin Li, and Philip S. Yu. 2021. HTCInfoMax: A global model for hier-
archical text classification via information maximization. In Proceedings of the 2021 Conference of the North Ameri-
can Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL-HLT 2021, On-
line, June 6-11, 2021, Kristina Toutanova, Anna Rumshisky, Luke Zettlemoyer, Dilek Hakkani-Tür, Iz Beltagy, Steven
Bethard, Ryan Cotterell, Tanmoy Chakraborty, and Yichao Zhou (Eds.). Association for Computational Linguistics,
3259–3265. http://dx.doi.org/10.18653/v1/2021.naacl-main.260
[176] Di Jin, Zhijing Jin, Joey Tianyi Zhou, and Peter Szolovits. 2020. Is BERT really robust? A strong baseline for natural
language attack on text classification and entailment. In The Thirty-Fourth AAAI Conference on Artificial Intelligence,
AAAI 2020, The Thirty-Second Innovative Applications of Artificial Intelligence Conference, IAAI 2020, The Tenth AAAI
Symposium on Educational Advances in Artificial Intelligence, EAAI 2020, New York, NY, USA, February 7-12, 2020.
8018–8025. https://aaai.org/ojs/index.php/AAAI/article/view/6311.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:36 Q. Li et al.

[177] Chen Li, Xutan Peng, Hao Peng, Jianxin Li, and Lihong Wang. 2021. TextGTL: Graph-based transductive learning
for semi-supervised TextClassification via structure-sensitive interpolation. In IJCAI 2021. ijcai.org.
[178] Takeru Miyato, Shin-ichi Maeda, Masanori Koyama, and Shin Ishii. 2017. Virtual adversarial training: A regular-
ization method for supervised and semi-supervised learning. CoRR abs/1704.03976 (2017). arXiv:1704.03976 http:
//arxiv.org/abs/1704.03976.
[179] Geoffrey E. Hinton, Alex Krizhevsky, and Sida D. Wang. 2011. Transforming auto-encoders. In Proc. ICANN, 2011,
Timo Honkela, Włodzisław Duch, Mark Girolami, and Samuel Kaski (Eds.). Springer Berlin, Berlin, 44–51.
[180] S. Hochreiter and J. Schmidhuber. 1997. Long short-term memory. Neural Computation 9, 8 (1997), 1735–1780.
[181] Samuel R. Bowman, Gabor Angeli, Christopher Potts, and Christopher D. Manning. 2015. A large annotated corpus
for learning natural language inference. In Proc. EMNLP, 2015. 632–642. http://dx.doi.org/10.18653/v1/d15-1075
[182] Zhiguo Wang, Wael Hamza, and Radu Florian. 2017. Bilateral multi-perspective matching for natural language sen-
tences. In Proc. IJCAI, 2017. 4144–4150. http://dx.doi.org/10.24963/ijcai.2017/579
[183] Rie Johnson and Tong Zhang. 2015. Semi-supervised convolutional neural networks for text categorization via re-
gion embedding. In Proc. NeurIPS, 2015. 919–927. http://papers.nips.cc/paper/5849-semi-supervised-convolutional-
neural-networks-for-text-categorization-via-region-embedding.
[184] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Identity mappings in deep residual networks. In
Proc. ECCV, 2016. 630–645. http://dx.doi.org/10.1007/978-3-319-46493-0_38
[185] Issam Bazzi. 2002. Modelling out-of-vocabulary words for robust speech recognition. Ph.D. Dissertation. Massachusetts
Institute of Technology.
[186] Huy Nguyen and Minh-Le Nguyen. 2017. A deep neural architecture for sentence-level sentiment classification in
Twitter social networking. In Proc. PACLING, 2017. 15–27. http://dx.doi.org/10.1007/978-981-10-8438-6_2
[187] Benjamin Adams and Grant McKenzie. 2018. Crowdsourcing the character of a place: Character-level convolutional
networks for multilingual geographic text classification. Trans. GIS 22, 2 (2018), 394–408. http://dx.doi.org/10.1111/
tgis.12317
[188] Zhuang Chen and Tieyun Qian. 2019. Transfer capsule network for aspect level sentiment classification. In Proc. ACL,
2019. 547–556. http://dx.doi.org/10.18653/v1/p19-1052
[189] Wei Xue, Wubai Zhou, Tao Li, and Qing Wang. 2017. MTNA: A neural multi-task model for aspect category classi-
fication and aspect term extraction on restaurant reviews. In Proc. IJCNLP, 2017. 151–156. https://www.aclweb.org/
anthology/I17-2026/.
[190] Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2015. Neural machine translation by jointly learning to
align and translate. In Proc. ICLR, 2015. http://arxiv.org/abs/1409.0473.
[191] Zikun Hu, Xiang Li, Cunchao Tu, Zhiyuan Liu, and Maosong Sun. 2018. Few-shot charge prediction with discrimi-
native legal attributes. In Proc. COLING, 2018. 487–498. https://www.aclweb.org/anthology/C18-1041/.
[192] Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and
Illia Polosukhin. 2017. Attention is all you need. In Proc. NeurIPS, 2017. 5998–6008. http://papers.nips.cc/paper/7181-
attention-is-all-you-need.
[193] Yukun Ma, Haiyun Peng, and Erik Cambria. 2018. Targeted aspect-based sentiment analysis via embedding com-
monsense knowledge into an attentive LSTM. In Proc. AAAI, 2018. 5876–5883. https://www.aaai.org/ocs/index.php/
AAAI/AAAI18/paper/view/16541.
[194] Yequan Wang, Minlie Huang, Xiaoyan Zhu, and Li Zhao. 2016. Attention-based LSTM for aspect-level sentiment
classification. In Proc. EMNLP, 2016. 606–615. http://dx.doi.org/10.18653/v1/d16-1058
[195] Feifan Fan, Yansong Feng, and Dongyan Zhao. 2018. Multi-grained attention network for aspect-level sentiment
classification. In Proc. EMNLP, 2018. 3433–3442. http://dx.doi.org/10.18653/v1/d18-1380
[196] Ming Tan, Cicero dos Santos, Bing Xiang, and Bowen Zhou. 2016. Improved representation learning for question
answer matching. In Proc. ACL, 2016. Association for Computational Linguistics, Berlin, Germany, 464–473. http:
//dx.doi.org/10.18653/v1/P16-1044
[197] Cícero Nogueira dos Santos, Ming Tan, Bing Xiang, and Bowen Zhou. 2016. Attentive pooling networks. CoRR
abs/1602.03609 (2016). arXiv:1602.03609 http://arxiv.org/abs/1602.03609.
[198] Xipeng Qiu, Tianxiang Sun, Yige Xu, Yunfan Shao, Ning Dai, and Xuanjing Huang. 2020. Pre-trained models for nat-
ural language processing: A survey. CoRR abs/2003.08271 (2020). arXiv:2003.08271 https://arxiv.org/abs/2003.08271.
[199] Alec Radford. 2018. Improving language understanding by generative pre-training.
[200] Zihang Dai, Zhilin Yang, Yiming Yang, Jaime G. Carbonell, Quoc Viet Le, and Ruslan Salakhutdinov. 2019.
Transformer-XL: Attentive language models beyond a fixed-length context. In Proc. ACL, 2019. 2978–2988. http:
//dx.doi.org/10.18653/v1/p19-1285
[201] Ganesh Jawahar, Benoît Sagot, and Djamé Seddah. 2019. What does BERT learn about the structure of language? In
Proc. ACL, 2019. 3651–3657. http://dx.doi.org/10.18653/v1/p19-1356
[202] Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Veselin
Stoyanov, and Luke Zettlemoyer. 2020. BART: Denoising sequence-to-sequence pre-training for natural language

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:37

generation, translation, and comprehension. In Proceedings of the 58th Annual Meeting of the Association for Compu-
tational Linguistics, ACL 2020, Online, July 5-10, 2020. 7871–7880. http://dx.doi.org/10.18653/v1/2020.acl-main.703
[203] Mandar Joshi, Danqi Chen, Yinhan Liu, Daniel S. Weld, Luke Zettlemoyer, and Omer Levy. 2020. SpanBERT:
Improving pre-training by representing and predicting spans. Trans. Assoc. Comput. Linguistics 8 (2020), 64–77.
https://transacl.org/ojs/index.php/tacl/article/view/1853.
[204] Yu Sun, Shuohuan Wang, Yu-Kun Li, Shikun Feng, Xuyi Chen, Han Zhang, Xin Tian, Danxiang Zhu, Hao Tian,
and Hua Wu. 2019. ERNIE: Enhanced representation through knowledge integration. CoRR abs/1904.09223 (2019).
arXiv:1904.09223 http://arxiv.org/abs/1904.09223.
[205] Michaël Defferrard, Xavier Bresson, and Pierre Vandergheynst. 2016. Convolutional neural networks on graphs with
fast localized spectral filtering. In Proc. NeurIPS, 2016. 3837–3845. http://papers.nips.cc/paper/6081-convolutional-
neural-networks-on-graphs-with-fast-localized-spectral-filtering.
[206] Hao Peng, Ruitong Zhang, Yingtong Dou, Renyu Yang, Jingyi Zhang, and Philip S. Yu. 2021. Reinforced neighborhood
selection guided multi-relational graph neural networks. arXiv preprint arXiv:2104.07886 (2021).
[207] Hao Peng, Renyu Yang, Zheng Wang, Jianxin Li, Lifang He, Philip Yu, Albert Zomaya, and Raj Ranjan. 2021. Lime:
Low-cost incremental learning for dynamic heterogeneous information networks. IEEE Trans. Comput. (2021).
[208] Jianxin Li, Hao Peng, Yuwei Cao, Yingtong Dou, Hekai Zhang, Philip Yu, and Lifang He. 2021. Higher-order attribute-
enhancing heterogeneous graph neural networks. IEEE Transactions on Knowledge and Data Engineering (2021).
[209] Diego Marcheggiani and Ivan Titov. 2017. Encoding sentences with graph convolutional networks for semantic role
labeling. In Proc. EMNLP, 2017. 1506–1515. http://dx.doi.org/10.18653/v1/d17-1159
[210] Yifu Li, Ran Jin, and Yuan Luo. 2019. Classifying relations in clinical narratives using segment graph convolutional
and recurrent neural networks (Seg-GCRNs). JAMIA 26, 3 (2019), 262–268. http://dx.doi.org/10.1093/jamia/ocy157
[211] Joost Bastings, Ivan Titov, Wilker Aziz, Diego Marcheggiani, and Khalil Sima’an. 2017. Graph convolutional encoders
for syntax-aware neural machine translation. In Proc. EMNLP, 2017. 1957–1967. http://dx.doi.org/10.18653/v1/d17-
1209
[212] Petar Velickovic, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Liò, and Yoshua Bengio. 2018. Graph
attention networks. In Proc. ICLR, 2018. https://openreview.net/forum?id=rJXMpikCZ.
[213] Linmei Hu, Tianchi Yang, Chuan Shi, Houye Ji, and Xiaoli Li. 2019. Heterogeneous graph attention networks for
semi-supervised short text classification. In Proc. EMNLP, 2019. 4820–4829. http://dx.doi.org/10.18653/v1/D19-1488
[214] Zhongyang Li, Xiao Ding, and Ting Liu. 2018. Constructing narrative event evolutionary graph for script event
prediction. In Proc. IJCAI, 2018. 4201–4207. http://dx.doi.org/10.24963/ijcai.2018/584
[215] Jane Bromley, Isabelle Guyon, Yann LeCun, Eduard Säckinger, and Roopak Shah. 1993. Signature verification using
a siamese time delay neural network. In Proc. NeurIPS, 1993]. 737–744. http://papers.nips.cc/paper/769-signature-
verification-using-a-siamese-time-delay-neural-network.
[216] Jonas Mueller and Aditya Thyagarajan. 2016. Siamese recurrent architectures for learning sentence similarity. In
Proc. AAAI, 2016. 2786–2792. http://www.aaai.org/ocs/index.php/AAAI/AAAI16/paper/view/12195.
[217] Jayadeva, Himanshu Pant, Mayank Sharma, and Sumit Soman. 2019. Twin neural networks for the classification of
large unbalanced datasets. Neurocomputing 343 (2019), 34–49. http://dx.doi.org/10.1016/j.neucom.2018.07.089 Learn-
ing in the Presence of Class Imbalance and Concept Drift.
[218] Takeru Miyato, Shin ichi Maeda, Masanori Koyama, Ken Nakae, and Shin Ishii. 2015. Distributional Smoothing with
Virtual Adversarial Training. (2015). arXiv:stat.ML/1507.00677.
[219] Tianyang Zhang, Minlie Huang, and Li Zhao. 2018. Learning structured representation for text classification via
reinforcement learning. In Proc. AAAI, 2018. 6053–6060. https://www.aaai.org/ocs/index.php/AAAI/AAAI18/paper/
view/16537.
[220] Jason Weston, Sumit Chopra, and Antoine Bordes. 2015. Memory Networks. (2015). arXiv:cs.AI/1410.3916.
[221] Xin Li and Wai Lam. 2017. Deep multi-task learning for aspect term extraction with memory interaction. In Proc.
EMNLP, 2017. 2886–2892. http://dx.doi.org/10.18653/v1/d17-1310
[222] Chenlin Shen, Changlong Sun, Jingjing Wang, Yangyang Kang, Shoushan Li, Xiaozhong Liu, Luo Si, Min Zhang, and
Guodong Zhou. 2018. Sentiment classification towards question-answering with hierarchical matching network. In
Proc. EMNLP, 2018. 3654–3663. http://dx.doi.org/10.18653/v1/d18-1401
[223] Xiao Ding, Kuo Liao, Ting Liu, Zhongyang Li, and Junwen Duan. 2019. Event representation learning enhanced with
external commonsense knowledge. In Proc. EMNLP, 2019. 4893–4902. http://dx.doi.org/10.18653/v1/D19-1495
[224] Yazhou Zhang, Dawei Song, Peng Zhang, Xiang Li, and Panpan Wang. 2019. A quantum-inspired sentiment repre-
sentation model for Twitter sentiment analysis. Applied Intelligence (2019).
[225] Ofir Zafrir, Guy Boudoukh, Peter Izsak, and Moshe Wasserblat. 2019. Q8BERT: Quantized 8Bit BERT. In Fifth Work-
shop on Energy Efficient Machine Learning and Cognitive Computing - NeurIPS Edition, EMC2@NeurIPS 2019, Vancou-
ver, Canada, December 13, 2019. 36–39. http://dx.doi.org/10.1109/EMC2-NIPS53020.2019.00016
[226] 2002. MR Corpus. http://www.cs.cornell.edu/people/pabo/movie-review-data/. (2002).

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:38 Q. Li et al.

[227] 2013. SST Corpus. http://nlp.stanford.edu/sentiment. (2013).

[228] Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Christopher D. Manning, Andrew Ng, and Christopher Potts.
2013. Recursive deep models for semantic compositionality over a sentiment treebank. In Proc. EMNLP, 2013. Associ-
ation for Computational Linguistics, Seattle, Washington, USA, 1631–1642. https://www.aclweb.org/anthology/D13-
1170.
[229] 2005. MPQA Corpus. http://www.cs.pitt.edu/mpqa/. (2005).
[230] Qiming Diao, Minghui Qiu, Chao-Yuan Wu, Alexander J. Smola, Jing Jiang, and Chong Wang. 2014. Jointly modeling
aspects, ratings and sentiments for movie recommendation (JMARS). In Proc. ACM SIGKDD, 2014. 193–202. http:
//dx.doi.org/10.1145/2623330.2623758
[231] Duyu Tang, Bing Qin, and Ting Liu. 2015. Document modeling with gated recurrent neural network for sentiment
classification. In Proc. EMNLP, 2015. 1422–1432. http://dx.doi.org/10.18653/v1/d15-1167
[232] 2015. Amazon Review Corpus. https://www.kaggle.com/datafiniti/consumer-reviews-of-amazon-products. (2015).
[233] 2013. Twitter Corpus. https://www.cs.york.ac.uk/semeval-2013/task2/. (2013).
[234] 2004. AG Corpus. http://www.di.unipi.it/~gulli/AG_corpus_of_news_articles.html. (2004).
[235] 2007. Reuters Corpus. https://www.cs.umb.edu/~smimarog/textmining/datasets/. (2007).
[236] Canhui Wang, Min Zhang, Shaoping Ma, and Liyun Ru. 2008. Automatic online news issue construction in web
environment. In Proc. WWW, 2008. 457–466. http://dx.doi.org/10.1145/1367497.1367560
[237] Ximing Li, Changchun Li, Jinjin Chi, Jihong Ouyang, and Chenliang Li. 2018. Dataless text classification: A topic
modeling approach with document manifold. In Proceedings of the 27th ACM International Conference on Information
and Knowledge Management, CIKM 2018, Torino, Italy, October 22-26, 2018. 973–982. http://dx.doi.org/10.1145/3269206.
3271671
[238] Jens Lehmann, Robert Isele, Max Jakob, Anja Jentzsch, Dimitris Kontokostas, Pablo N. Mendes, Sebastian Hellmann,
Mohamed Morsey, Patrick van Kleef, Sören Auer, and Christian Bizer. 2015. DBpedia - A large-scale, multilingual
knowledge base extracted from Wikipedia. Semantic Web 6, 2 (2015), 167–195. http://dx.doi.org/10.3233/SW-140134
[239] 2015. Ohsumed Corpus. http://davis.wpi.edu/xmdv/datasets/ohsumed.html. (2015).
[240] 2019. EUR-Lex Corpus. http://www.ke.tu-darmstadt.de/resources/eurlex/eurlex.html. (2019).
[241] 2016. Amazon670K Corpus. http://manikvarma.org/downloads/XC/XMLRepository.html. (2016).
[242] Jianhua Yin and Jianyong Wang. 2014. A dirichlet multinomial mixture model-based approach for short text cluster-
ing. In The 20th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD’14, New York,
NY, USA - August 24 - 27, 2014. 233–242. http://dx.doi.org/10.1145/2623330.2623715
[243] J. Chen, Z. Gong, W. Wang, W. Liu, M. Yang, and C. Wang. 2020. TAM: Targeted analysis model with reinforcement
learning on short texts. IEEE Transactions on Neural Networks and Learning Systems (2020), 1–10. http://dx.doi.org/
10.1109/TNNLS.2020.3009247
[244] Fang Wang, Zhongyuan Wang, Zhoujun Li, and Ji-Rong Wen. 2014. Concept-based short text classification and
ranking. In Proc. CIKM, 2014. 1069–1078. http://dx.doi.org/10.1145/2661829.2662067
[245] 2015. Fudan Corpus. www.datatang.com/data/44139and43543. (2015).
[246] Pranav Rajpurkar, Jian Zhang, Konstantin Lopyrev, and Percy Liang. 2016. SQuAD: 100, 000+ questions for machine
comprehension of text. In Proc. EMNLP, 2016. 2383–2392. http://dx.doi.org/10.18653/v1/d16-1264
[247] Xuchen Yao, Benjamin Van Durme, Chris Callison-Burch, and Peter Clark. 2013. Answer extraction as sequence
tagging with tree edit distance. In Proc. NAACL, 2013. 858–867. https://www.aclweb.org/anthology/N13-1106/.
[248] 2002. TREC Corpus. https://cogcomp.seas.upenn.edu/Data/QA/QC/. (2002).
[249] Yi Yang, Wen-tau Yih, and Christopher Meek. 2015. WikiQA: A challenge dataset for open-domain question answer-
ing. In Proc. EMNLP, 2015. 2013–2018. http://dx.doi.org/10.18653/v1/d15-1237
[250] Bo Pang and Lillian Lee. 2004. A sentimental education: Sentiment analysis using subjectivity summarization based
on minimum cuts. In Proc. ACL, 2004. Barcelona, Spain, 271–278. http://dx.doi.org/10.3115/1218955.1218990
[251] Minqing Hu and Bing Liu. 2004. Mining and summarizing customer reviews. In Proc. ACM SIGKDD, 2004. 168–177.
http://dx.doi.org/10.1145/1014052.1014073
[252] 2017. Reuters Corpus. https://martin-thoma.com/nlp-reuters. (2017).
[253] Jaeyoung Kim, Sion Jang, Eunjeong L. Park, and Sungchul Choi. 2020. Text classification using capsules. Neurocom-
puting 376 (2020), 214–221. http://dx.doi.org/10.1016/j.neucom.2019.10.033
[254] 2020. Reuters10 Corpus. http://www.nltk.org/book/ch02.html. (2020).
[255] David D. Lewis, Yiming Yang, Tony G. Rose, and Fan Li. 2004. RCV1: A new benchmark collection for text catego-
rization research. J. Mach. Learn. Res. 5 (2004), 361–397. http://jmlr.org/papers/volume5/lewis04a/lewis04a.pdf.
[256] 2004. RCV1-V2 Corpus. http://www.ai.mit.edu/projects/jmlr/papers/volume5/lewis04a/lyrl2004_rcv1v2_README.
htm. (2004).
[257] Bo Pang and Lillian Lee. 2005. Seeing stars: Exploiting class relationships for sentiment categorization with respect
to rating scales. In Proc. ACL, 2005. 115–124. https://www.aclweb.org/anthology/P05-1015/.

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:39

[258] Janyce Wiebe, Theresa Wilson, and Claire Cardie. 2005. Annotating expressions of opinions and emotions in lan-
guage. Language Resources and Evaluation 39, 2–3 (2005), 165–210. http://dx.doi.org/10.1007/s10579-005-7880-9
[259] Mike Thelwall, Kevan Buckley, and Georgios Paltoglou. 2012. Sentiment strength detection for the social web. J.
Assoc. Inf. Sci. Technol. 63, 1 (2012), 163–173. http://dx.doi.org/10.1002/asi.21662
[260] Preslav Nakov, Alan Ritter, Sara Rosenthal, Fabrizio Sebastiani, and Veselin Stoyanov. 2016. SemEval-2016 task 4:
Sentiment analysis in Twitter. In Proc. SemEval, 2016.
[261] Zhiyong Lu. 2011. PubMed and beyond: A survey of web tools for searching biomedical literature. Database 2011
(2011). http://dx.doi.org/10.1093/database/baq036
[262] Tri Nguyen, Mir Rosenberg, Xia Song, Jianfeng Gao, Saurabh Tiwary, Rangan Majumder, and Li Deng. 2016. MS
MARCO: A human generated machine reading comprehension dataset. In Proc. NeurIPS, 2016. http://ceur-ws.org/Vol-
1773/CoCoNIPS_2016_paper9.pdf.
[263] https://data.quora.com/First-Quora-Dataset-Release-QuestionPairs. ([n. d.]).
[264] Pranav Rajpurkar, Robin Jia, and Percy Liang. 2018. Know what you don’t know: Unanswerable questions for SQuAD.
In Proc. ACL, 2018. 784–789. http://dx.doi.org/10.18653/v1/P18-2124
[265] Adina Williams, Nikita Nangia, and Samuel R. Bowman. 2018. A broad-coverage challenge corpus for sentence un-
derstanding through inference. In Proc. NAACL, 2018. 1112–1122. http://dx.doi.org/10.18653/v1/n18-1101
[266] Marco Marelli, Luisa Bentivogli, Marco Baroni, Raffaella Bernardi, Stefano Menini, and Roberto Zamparelli. 2014.
SemEval-2014 task 1: Evaluation of compositional distributional semantic models on full sentences through semantic
relatedness and textual entailment. In Proc. SemEval, 2014. 1–8. http://dx.doi.org/10.3115/v1/s14-2001
[267] Bill Dolan, Chris Quirk, and Chris Brockett. 2004. Unsupervised construction of large paraphrase corpora: Exploiting
massively parallel news sources. In Proc. COLING, 2004. https://www.aclweb.org/anthology/C04-1051/.
[268] Daniel M. Cer, Mona T. Diab, Eneko Agirre, Iñigo Lopez-Gazpio, and Lucia Specia. 2017. SemEval-2017 task
1: Semantic textual similarity - multilingual and cross-lingual focused evaluation. CoRR abs/1708.00055 (2017).
arXiv:1708.00055 http://arxiv.org/abs/1708.00055.
[269] Ido Dagan, Oren Glickman, and Bernardo Magnini. 2005. The PASCAL recognising textual entailment challenge. In
Machine Learning Challenges, Evaluating Predictive Uncertainty, Visual Object Classification and Recognizing Textual
Entailment, First PASCAL Machine Learning Challenges Workshop, MLCW 2005, Southampton, UK, April 11-13, 2005,
Revised Selected Papers. 177–190. http://dx.doi.org/10.1007/11736790_9
[270] Tushar Khot, Ashish Sabharwal, and Peter Clark. 2018. SciTaiL: A textual entailment dataset from science question
answering. In Proceedings of the Thirty-Second AAAI Conference on Artificial Intelligence, (AAAI-18), the 30th Innova-
tive Applications of Artificial Intelligence (IAAI-18), and the 8th AAAI Symposium on Educational Advances in Artificial
Intelligence (EAAI-18), New Orleans, Louisiana, USA, February 2-7, 2018. 5189–5197. https://www.aaai.org/ocs/index.
php/AAAI/AAAI18/paper/view/17368.
[271] Kamran Kowsari, Donald E. Brown, Mojtaba Heidarysafa, Kiana Jafari Meimandi, Matthew S. Gerber, and Laura E.
Barnes. 2017. HDLTex: Hierarchical deep learning for text classification. In Proc. ICMLA, 2017. 364–371. http://dx.doi.
org/10.1109/ICMLA.2017.0-134
[272] 2018. AmazonCat-13K Corpus. https://drive.google.com/open?id=1VwHAbri6y6oh8lkpZ6sSY_b1FRNnCLFL. (2018).
[273] 2017. BlurbGenreCollection-EN Corpus. https://www.inf.uni-hamburg.de/en/inst/ab/lt/resources/data/blurb-genre-
collection.html. (2017).
[274] Iris Hendrickx, Su Nam Kim, Zornitsa Kozareva, Preslav Nakov, Diarmuid Ó Séaghdha, Sebastian Padó, Marco Pen-
nacchiotti, Lorenza Romano, and Stan Szpakowicz. 2009. SemEval-2010 task 8: Multi-way classification of semantic
relations between pairs of nominals. In Proc. NAACL, 2009. 94–99. https://www.aclweb.org/anthology/W09-2415/.
[275] Stephanie M. Strassel, Mark A. Przybocki, Kay Peterson, Zhiyi Song, and Kazuaki Maeda. 2008. Linguistic resources
and evaluation techniques for evaluation of cross-document automatic content extraction. In Proc. LREC, 2008. http:
//www.lrec-conf.org/proceedings/lrec2008/summaries/677.html.
[276] Yuhao Zhang, Victor Zhong, Danqi Chen, Gabor Angeli, and Christopher D. Manning. 2017. Position-aware attention
and supervised data improve slot filling. In Proc. EMNLP, 2017. 35–45. http://dx.doi.org/10.18653/v1/d17-1004
[277] Sebastian Riedel, Limin Yao, and Andrew McCallum. 2010. Modeling relations and their mentions without labeled
text. In Proc. ECML PKDD, 2010. 148–163. http://dx.doi.org/10.1007/978-3-642-15939-8_10
[278] 2019. FewRel Corpus. https://github.com/thunlp/FewRel. (2019).
[279] Seokhwan Kim, Luis Fernando D’Haro, Rafael E. Banchs, Jason D. Williams, and Matthew Henderson. 2016. The
fourth dialog state tracking challenge. In Proc. IWSDS, 2016. 435–449. http://dx.doi.org/10.1007/978-981-10-2585-3_36
[280] Jeremy Ang, Yang Liu, and Elizabeth Shriberg. 2005. Automatic dialog act segmentation and classification in multi-
party meetings. In Proc. ICASSP, 2005. 1061–1064. http://dx.doi.org/10.1109/ICASSP.2005.1415300
[281] Dan Jurafsky and Elizabeth Shriberg. 1997. Switchboard SWBD-DAMSL shallow-discourse-function annotation
coders manual. (01 1997).

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
31:40 Q. Li et al.

[282] Aliaksei Severyn and Alessandro Moschitti. 2015. Learning to rank short text pairs with convolutional deep neural
networks. In Proceedings of the 38th International ACM SIGIR Conference on Research and Development in Information
Retrieval, Santiago, Chile, August 9-13, 2015, Ricardo Baeza-Yates, Mounia Lalmas, Alistair Moffat, and Berthier A.
Ribeiro-Neto (Eds.). ACM, 373–382. http://dx.doi.org/10.1145/2766462.2767738
[283] Christopher D. Manning, Prabhakar Raghavan, and Hinrich Schütze. 2008. Introduction to Information Retrieval. Cam-
bridge University Press. http://dx.doi.org/10.1017/CBO9780511809071
[284] Tetsuji Nakagawa, Kentaro Inui, and Sadao Kurohashi. 2010. Dependency tree-based sentiment classification us-
ing CRFs with hidden variables. In Human Language Technologies: Conference of the North American Chapter of
the Association of Computational Linguistics, Proceedings, June 2-4, 2010, Los Angeles, California, USA. 786–794.
https://aclanthology.org/N10-1120/.
[285] Jeremy Howard and Sebastian Ruder. 2018. Universal language model fine-tuning for text classification. In Proc. ACL,
2018. 328–339. http://dx.doi.org/10.18653/v1/P18-1031
[286] Guoyin Wang, Chunyuan Li, Wenlin Wang, Yizhe Zhang, Dinghan Shen, Xinyuan Zhang, Ricardo Henao, and
Lawrence Carin. 2018. Joint embedding of words and labels for text classification. In Proc. ACL, 2018. 2321–2331.
http://dx.doi.org/10.18653/v1/P18-1216
[287] Xiaodong Liu, Pengcheng He, Weizhu Chen, and Jianfeng Gao. 2019. Multi-task deep neural networks for natural
language understanding. In Proc. ACL, 2019. 4487–4496. http://dx.doi.org/10.18653/v1/p19-1441
[288] Pushpankar Kumar Pushp and Muktabh Mayank Srivastava. 2017. Train once, test anywhere: Zero-shot learning for
text classification. CoRR abs/1712.05972 (2017). arXiv:1712.05972 http://arxiv.org/abs/1712.05972.
[289] Congzheng Song, Shanghang Zhang, Najmeh Sadoughi, Pengtao Xie, and Eric P. Xing. 2020. Generalized zero-shot
text classification for ICD coding. In Proc. IJCAI, 2020. 4018–4024. http://dx.doi.org/10.24963/ijcai.2020/556
[290] Ruiying Geng, Binhua Li, Yongbin Li, Xiaodan Zhu, Ping Jian, and Jian Sun. 2019. Induction networks for few-shot
text classification. In Proc. EMNLP, 2019. 3902–3911. http://dx.doi.org/10.18653/v1/D19-1403
[291] Shumin Deng, Ningyu Zhang, Zhanlin Sun, Jiaoyan Chen, and Huajun Chen. 2020. When low resource NLP meets
unsupervised language model: Meta-pretraining then meta-learning for few-shot text classification (student abstract).
In Proc. AAAI, 2020. 13773–13774. https://aaai.org/ojs/index.php/AAAI/article/view/7158.
[292] Ruiying Geng, Binhua Li, Yongbin Li, Jian Sun, and Xiaodan Zhu. 2020. Dynamic memory induction networks for
few-shot text classification. In Proc. ACL, 2020. 1087–1094. https://www.aclweb.org/anthology/2020.acl-main.102/.
[293] Kervy Rivas Rojas, Gina Bustamante, Arturo Oncevay, and Marco Antonio Sobrevilla Cabezudo. 2020. Efficient
strategies for hierarchical text classification: External knowledge and auxiliary tasks. In Proceedings of the 58th
Annual Meeting of the Association for Computational Linguistics, ACL 2020, Online, July 5–10, 2020. 2252–2257.
http://dx.doi.org/10.18653/v1/2020.acl-main.205
[294] Niloofer Shanavas, Hui Wang, Zhiwei Lin, and Glenn I. Hawe. 2021. Knowledge-driven graph similarity for text
classification. Int. J. Mach. Learn. Cybern. 12, 4 (2021), 1067–1081. http://dx.doi.org/10.1007/s13042-020-01221-4
[295] Yanchao Hao, Yuanzhe Zhang, Kang Liu, Shizhu He, Zhanyi Liu, Hua Wu, and Jun Zhao. 2017. An end-to-end model
for question answering over knowledge base with cross-attention combining global knowledge. In Proc. ACL, 2017.
221–231. http://dx.doi.org/10.18653/v1/P17-1021
[296] Rima Türker, Lei Zhang, Maria Koutraki, and Harald Sack. 2018. TECNE: Knowledge based text classification using
network embeddings. In Proc. EKAW, 2018. 53–56. http://ceur-ws.org/Vol-2262/ekaw-demo-18.pdf.
[297] Xin Liang, Dawei Cheng, Fangzhou Yang, Yifeng Luo, Weining Qian, and Aoying Zhou. 2020. F-HMTC: Detecting
financial events for investment decisions based on neural hierarchical multi-label text classification. In Proceedings
of the Twenty-Ninth International Joint Conference on Artificial Intelligence, IJCAI 2020. 4490–4496. http://dx.doi.org/
10.24963/ijcai.2020/619
[298] P. B. Shanthi, Shivani Modi, K. S. Hareesha, and Sampath Kumar. 2021. Classification and comparison of malignancy
detection of cervical cells based on nucleus and textural features in microscopic images of uterine cervix. Int. J.
Medical Eng. Informatics 13, 1 (2021), 1–13. http://dx.doi.org/10.1504/IJMEI.2021.111861
[299] Tianshi Wang, Li Liu, Naiwen Liu, Huaxiang Zhang, Long Zhang, and Shanshan Feng. 2020. A multi-label text
classification method via dynamic semantic representation model and deep neural network. Appl. Intell. 50, 8 (2020),
2339–2351. http://dx.doi.org/10.1007/s10489-020-01680-w
[300] Boyan Wang, Xuegang Hu, Pei-Pei Li, and Philip S. Yu. 2021. Cognitive structure learning model for hierarchical
multi-label text classification. Knowl. Based Syst. 218 (2021), 106876. http://dx.doi.org/10.1016/j.knosys.2021.106876
[301] Jinhua Du, Yan Huang, and Karo Moilanen. 2020. Pointing to select: A fast pointer-LSTM for long text classifica-
tion. In Proceedings of the 28th International Conference on Computational Linguistics, COLING 2020, Barcelona, Spain
(Online), December 8-13, 2020. 6184–6193. http://dx.doi.org/10.18653/v1/2020.coling-main.544
[302] Jie Du, Chi-Man Vong, and C. L. Philip Chen. 2021. Novel efficient RNN and LSTM-like architectures: Recurrent and
gated broad learning systems and their applications for text classification. IEEE Trans. Cybern. 51, 3 (2021), 1586–1597.
http://dx.doi.org/10.1109/TCYB.2020.2969705

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.
A Survey on Text Classification 31:41

[303] Teja Kanchinadam, Qian You, Keith Westpfahl, James Kim, Siva Gunda, Sebastian Seith, and Glenn Fung. 2021.
A simple yet brisk and efficient active learning platform for text classification. CoRR abs/2102.00426 (2021).
arXiv:2102.00426 https://arxiv.org/abs/2102.00426.
[304] Yichao Zhou, Jyun-Yu Jiang, Kai-Wei Chang, and Wei Wang. 2019. Learning to discriminate perturbations for block-
ing adversarial attacks in text classification. 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 2019,
Hong Kong, China, November 3-7, 2019. 4903–4912. http://dx.doi.org/10.18653/v1/D19-1496
[305] Ahmadreza Azizi, Ibrahim Asadullah Tahmid, Asim Waheed, Neal Mangaokar, Jiameng Pu, Mobin Javed, Chandan K.
Reddy, and Bimal Viswanath. 2021. T-Miner: A generative approach to defend against trojan attacks on DNN-based
text classification. CoRR abs/2103.04264 (2021). arXiv:2103.04264 https://arxiv.org/abs/2103.04264.

Received April 2021; revised August 2021; accepted October 2021

ACM Transactions on Intelligent Systems and Technology, Vol. 13, No. 2, Article 31. Publication date: April 2022.

View publication stats