Indonesian Journal of Electrical Engineering and Computer Science

Vol. 34, No. 2, May 2024, pp. 1258~1267

ISSN: 2502-4752, DOI: 10.11591/ijeecs.v34.i2.pp1258-1267  1258

Text classification on user feedback: a systematic literatures

review

Zuleaizal Sidek1,2, Sharifah Sakinah Syed Ahmad1, Yogan Jaya Kumar1, Noor Hasimah Ibrahim Teo3
1
Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka (UTeM), Melaka, Malaysia
2
Institut Tun Perak, Melaka, Malaysia
3
School of Computing, College of Computing, Informatics and Mathematics, Universiti Teknologi MARA (Melaka Branch),
Melaka, Malaysia

Article Info ABSTRACT

Article history: User feedback in text classification serves multiple purposes, including
refining models, enhancing datasets, adapting to user preferences,
Received Feb 6, 2024 identifying emerging topics, and evaluating system performance, all of
Revised Feb 19, 2024 which contribute to the creation of more effective and user-centric
Accepted Feb 22, 2024 classification systems. Many text classification techniques, including data
mining, machine learning, and deep learning approaches, have been
employed in previous literature, each making significant contributions to the
Keywords: field. This paper aims to contribute by guiding researchers seeking
commonly used classification techniques and evaluation metrics in text
Customer feedback processing. Additionally, it identifies the classification technique that
Customer reviews generates higher accuracy and works as a basis for researchers to synthesize
Literature review studies within their respective fields. Preferred reporting items for systematic
Text classification reviews and meta-analysis (PRISMA) methodology is adapted to
Textual data systematically review 28 current literatures on text classification on user
feedback. The results obtained are guided by four research questions; paper
distribution year, dataset source and size; evaluation metric and model
accuracy. The review has shown that support vector machines (SVM) are
frequently employed and consistently achieve high levels of accuracy as
high as 97.17% with various datasets used. The future direction of this work
could explore models that integrate sentiment analysis and natural language
understanding to more accurately capture nuanced user opinions and
preferences.
This is an open access article under the CC BY-SA license.

Corresponding Author:
Zuleaizal Sidek
Faculty of Information and Communication Technology, Universiti Teknikal Malaysia Melaka (UTeM)
76100, Durian Tunggal, Melaka, Malaysia
Email: [email protected]

1. INTRODUCTION
The number of people actively participating virtually on the internet, particularly social media, is
constantly rising, especially during the pandemic [1]. This has resulted in the rise of e-commerce, which
refers to the provision of services via the internet and allows for instant communication and trading, thus
cutting down on the amount of time needed to make a purchase or send an item. The internet has a vast
quantity of text-based content, posing significant challenges for manual classification [2]. The complexity of
handling various types of data, such as emails, patient reports derived from medical records, academic papers
with specific technological needs, and news articles categorized by different topics, has increased
significantly due to the large volume and numerous obstacles associated with identifying them as either

Journal homepage: http://ijeecs.iaescore.com

Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1259

malicious or safe. There exists a need for an automated technique that can effectively classify textual content
in a manner that is both straightforward and compatible with machines [3].
Customer feedback texts play a crucial role in facilitating a direct channel for consumers to
articulate their thoughts, sentiments, and experiences [4]. It has been proposed that users of social media
platforms are not only engaged in product reviews, but also play a variety of other roles, including those of
online activists, trolls, social critics, information seekers, and socialites. As a result, doing an appropriate
analysis of their feelings is of the utmost importance [5]. This feedback encompasses a wide range of
opinions, varying from highly positive recommendations to harsh criticisms, and serves as a valuable
resource for businesses, offering a wealth of knowledge. Through the process of detecting reoccurring issues
and client preferences, organizations can have the opportunity to modify their products, thus improving their
overall attractiveness and functionality. The customer feedback may consist of complaints, reports, or
inquiries about the services rendered by the company. Complaints should not be regarded as an
inconvenience that needs to be evaded, dismissed, or concealed, as they have the potential to produce novel
ideas for assessing and enhancing both products and services [6]. If a client’s complaint isn’t resolved to their
satisfaction, they may decide to stop being a customer immediately. Therefore, businesses should motivate
clients to provide feedback on services, particularly in the event of a negative experience.
Customer reviews often utilize a five-point rating scale to gauge customer satisfaction, along with
textual input for additional comments or feedback. The additional remarks consist of text messages that
possess significant utility, demanding the implementation of an intelligent analysis mechanism. This
mechanism should be capable of analyzing these messages through the utilization of several techniques,
including natural language processing, ontology [7], and classification approaches [8]. Putting the importance
of text classification based on consumer feedback into perspective requires thinking about the massive
amount of data created on a worldwide scale. A recent study estimates yearly global data production at an
astonishing 64.2 zettabytes [9]. A substantial portion of this data is comprised of textual customer feedback,
spanning multiple languages, cultures, and industries. This exponential growth in data volume underscores
the urgent need for sophisticated techniques to extract actionable insights.
The problem of text classification refers to the task of assigning text documents to predetermined
classes or groups depending on the content of the processing texts, utilizing data mining tools and
approaches. Text classification applications have proven to be highly effective across several domains,
serving diverse objectives [10], [11]. In the context of text classification, numerous researchers have
contributed significant improvements in their studies. The subject matter encompasses the classification of
textual datasets across domains using machine learning (ML) and deep learning (DL), as well as the
utilization of statistical techniques.
Improve retrieval performance and speed up the process of extracting useful information from vast
amounts of text data with the help of text classification technology, which is able to automate and simplify
complex text information [12]. There are two main categories of textual representation: factual and opinion-
based. Facts can be interpreted as objective representations of items, events, and their attributes, whereas
opinions are subjective statements that reflect individuals’ sentiments towards these entities or occurrences [13].
Concerning understanding consumer feedback or reviews, sentiment analysis is becoming a viable technique
for businesses [14]. In particular, customer reviews that include people’s attitudes toward products are being
analysed with sentiment analysis algorithms [15]. Reviews, comments, surveys, online resources, and social
media can all be leveraged for better customer relationship management through sentiment analysis.
The identification of intent plays a vital role in dialogue systems since it enables the system to
comprehend the user’s desired actions [16]. Therefore, this paper presents a study on text classification
methods used for customer feedback texts, focusing on the methods and techniques used in the classification
process. This paper will address the following research questions:
RQ1: what is the distribution year, author, domain application, classification types related to the text
classification and also sources and sizes of the dataset used?
RQ2: what are the sources and sizes of datasets used in the reviewed studies?
RQ3: what are the most common performance evaluation metrics used?
RQ4: what are the techniques/approaches gives higher accuracy?
The subsequent sections of this paper are structured as follows. Section 2 presents the methodology
employed in conducting a systematic review. Section 3 provides an analysis of the relevant literature, findings
and discussions. The investigation is ultimately ended in section 4 with the conclusion and future work.

2. METHOD
The approach of conducting a systematic literature review (SLR) has been chosen as the research
method for this research. The SLR is a protocol-based method that was developed in response to the research
questions to locate, select, and evaluate the literature that is the most pertinent [17]. There are a few key
Text classification on user feedback: a systematic literatures review (Zuleaizal Sidek)
1260  ISSN: 2502-4752

principles such as coverage and focus were taken into consideration when deciding whether to use SLR as a
method of review. The SLR procedure, which is suggested in the literature [18], [19], is followed by this study
to give a more objective and comprehensive literature review. As can be seen in Figure 1, the SLR process is
broken up into three key phases: planning, performing the review, and reporting the results of the review.

Figure 1. Strategic literature reviews process

This research employed the preferred reporting items for systematic reviews and meta-analysis
(PRISMA) technique to perform a systematic review (SR) for this work. The PRISMA is a standardized
collection of essential elements that serve as a foundation for the construction and organization of SRs and
other forms of meta-analyses. Therefore, the reviewing process is comprised of five stages, namely planning,
searching, filtering, eligibility and reviewing. It is grounded in empirical data and provides a framework for
ensuring the quality and comprehensiveness of these research efforts. The details of these procedures, as they
pertain to our research investigation, are going to be covered in the parts that follow.

2.1. Planning phase

2.1.1. Defining of research question
This SR is an organized approach that includes the extent of the research that was evaluated by
classifying and analyzing the various publications that are already connected to this topic. Determining the
research questions that will be asked to correctly describe the percentage of previously published works is the
first step. When we read works that are related to one another, we can gain many insights that can later assist
scholars in developing new concepts. Table 1 contains a description of the research questions that were
utilized in our SR.

Table 1. Research question description

Research question Description
RQ1: what is the distribution year, author, domain The response to RQ1 facilitates the identification of the time,
application, classification types related to the text domain of application, classification type, and authors of the
classification and also sources and sizes of the dataset used? conducted research studies.
RQ2: what are the sources and sizes of datasets used in the The response to RQ2 enables researchers to gain an understanding
reviewed studies? of the biases or limitations inherent in the datasets.
RQ3: what are the most common performance evaluation The response to RQ3 outlines the metrics employed for assessing
metrics used? the efficacy of the developed techniques as the key findings of the
reviewed studies.
RQ4: what are the techniques/approaches gives higher The response to RQ4 is to identify suitable tools and techniques
accuracy? employed by the latest applications.

This research employed the PRISMA technique to perform a SR for this work. The PRISMA is a
standardized collection of essential elements that serve as a foundation for the construction and organization
of SRs and other forms of meta-analyses. Therefore, the reviewing process is comprised of five stages,
namely planning, searching, filtering, eligibility and reviewing. It is grounded in empirical data and provides
a framework for ensuring the quality and comprehensiveness of these research efforts.

Indonesian J Elec Eng & Comp Sci, Vol. 34, No. 2, May 2024: 1258-1267
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1261

2.1.2. Inclusion and exclusion criteria

We employ a set of inclusion criteria (IC) and exclusion criteria (EC) to identify whether papers are
relevant to our search in Table 2. This allows us to refine the results of our search. Studies that do not address
EC are disregarded, and an additional filtering technique is utilized to identify articles that meet our setting.

Table 2. Research question description

Inclusion criteria Exclusion criteria
Literature is published within the timeframe from 2019 to 2023. Duplicated literatures
Literature must satisfy at least one of the specified search terms Research papers that are not written in English
Literature should be published in reputable academic journals, Literature that is not relevant to the topic.
conferences, or academic magazines.
The search is conducted leveraging the title, abstract, and complete text.

The preliminary evaluation is comprised of the following three IC stages:

i) The following phase relies on the abstracts and entails eliminating important findings based on the
information and keywords identified in the abstracts. Only the abstracts of articles approved by the IC
were kept for further processing.
ii) Step based on the full text: we reject the results that neglect to address or describe the research terms in
Table 3, such as publications that only represent minor elements of the search terms contained in their
abstracts. This is the step that we take after the step based on the full text.
iii) Phase based on quality analysis: we assess the quality of the remaining results, deleting any of them that
do not meet any of the following criteria:
C1: the publications present a thorough method for categorizing customer feedback.
C2: the publication details the technical aspects of how the suggested solution would be implemented.
C3: the publication cites several other studies.

2.2. Searching phase

There was a total of 484 papers found during the first phase of the search that had search phrases
included in the title and had the potential to be the most relevant. The search queries that were utilized for
this study are presented in Table 3. We considered the research presented at conferences, in journals, as part
of theses, and in reports that were published between 2019 and 2023.

Table 3. Searching term

Search ID Search term
S1 Text classification AND customer feedback
S2 Text classification AND customer reviews
S3 Intent classification AND customer reviews
S4 Sentiment analysis AND customer reviews
S5 Sentiment analysis AND customer feedback

The proposed approach comprises two main stages: i) analyzing search phrases from previous study
questions to create a list of keywords, and ii) formulating queries using Boolean operators AND/OR to find
and gather all relevant results. The first point generating a list of keywords by extracting search terms from
previous research queries. The initial phase in conducting our SR was to identify the various data sets that
would be used. As can be seen in Table 4, searches have been conducted using a variety of academic
databases, digital libraries, and search engines, including academic and open-access options. The next step is
to define the techniques that will be used to examine the academic and technical papers that these searches
yielded to locate publications that are pertinent to our context.

Table 4. Literature database

Database URL
Scopus digital library www.scopus.com
IEEE Xplore digital library https://ieeexplore.ieee.org/Xplore/home.jsp
ScienceDirect digital library https://www.sciencedirect.com/
ResearchGate https://researchgate.com

Text classification on user feedback: a systematic literatures review (Zuleaizal Sidek)

1262  ISSN: 2502-4752

2.2.1. Filtering
Relevant literature was selected for this review by following the planning, searching, and filtering
steps outlined in the PRISMA flow diagram. An illustration of the systematic search approach that will be
utilized can be found in Figure 1. During the preliminary stage, we were successful in acquiring 484 records.
We were left with 370 original results after we eliminated all the duplicate literature. Following that, criteria
for filtering based on the title and abstract were applied to these 370, and therefore, to 134 after this round of
elimination.

2.2.2. Eligibility
By including further criteria for eligibility outlined in the complete article, we were ultimately able
to obtain a total of 28 relevant research studies. A thorough evaluation was conducted on a collection of 28
research publications to extract the findings that will be elaborated upon in the subsequent section. Out of the
28 publications that through the review process, 16 of them employed machine-learning approaches for text
classification. Among these, 9 papers metrics DL methods, while 3 studies utilized statistical analysis.
The remaining publications employed a combination of fuzzy logic and the Ernie approach. Out of a total of 28
published literature, many of the studies employ ML approaches for text classification.

3. RESULTS AND DISCUSSION

This section discusses the most important findings that emerged from our comprehensive literature
reviews were based on primary research papers published between 2019 and 2023. This part focuses on the
findings of the review related to the research questions that were previously established. It has a total of
28 papers that were systematically selected and relevant to the domain of text classification in the context of
customer feedback and reviews. These discussions assist researchers in gaining information concerning
recent literature on the topic. It provides information on the methods employed, the techniques used for
evaluating text classification, and other specific tools employed in the research process.
RQ1: what is the distribution year, author, domain application, and classification types related to the
text classification? The literature distribution result is shown in Figure 2. Based on the reviews, most of the
literature was published in the year 2022, with a total count of 14 publications. In the year 2023, there are a
total of eight publications. From 2019 to 2021, there were a total of six literatures, with one literature in
2019, three in 2020, and two in 2021 as depicted in Figure 2(a). It can be observed that in the year 2022, a
significant number of academic researchers are engaged in the task of text classification, specifically
focusing on customer feedback and reviews. Out of the 28 selected literature, more than half of them
(16 sources) are utilizing ML as the primary methodology for researching the text classification of customer
feedback and review. The remaining publications utilize DL techniques, as well as a combination of ML and
statistical analysis methodologies as in Figure 2(b).

(a) (b)

Figure 2. Literature distribution based on (a) year of publication and (b) classification type

RQ2: what are the sources and sizes of datasets used in the reviewed studies? The datasets were
obtained from social media platforms such as Twitter, and other online platforms, as indicated in Table 5.
Most of the literature utilizes datasets consisting of Amazon customer reviews and tweets from the social
media platform Twitter. Three different studies have utilized TripAdvisor reviews, whereas two studies used
ATIS as datasets for their research. The rest of the literature uses several textual datasets for their analyses.

Indonesian J Elec Eng & Comp Sci, Vol. 34, No. 2, May 2024: 1258-1267
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1263

Table 5. List of literature and dataset sources

Literature Sources
[20]−[24] Amazon
[20], [21], [25]−[27] Twitter
[28], [29] TripAdvisor
[30], [31] ATIS
[22], [32]−[43] Others

Due to the large number of individuals who use it and the diversity of intriguing topics on which
they express their thoughts, Twitter has become the most popular social media network. In addition, scraping
the public live tweets from Twitter is made simpler when the Twitter API is used. Since it is the largest
online purchasing platform in the world, Amazon.com has become an increasingly popular source of
customer feedback for researchers conducting sentiment analysis. There are several of the listed literature,
such as [20]−[22], that believe that the quantity of datasets influences the accuracy of classification. When
common approaches that involve ML, one crucial consideration is the size of the dataset. According to what
is found in the research literature, the quality of model learning improves and the accuracy of the
generalization phase increases as the dataset size is increased. Because the dimensions of datasets may vary
from one research project to the next, we decided to include dataset size as one of the parameters in our
systematic reviews. The dataset sizes are categorized into five intervals, commencing with datasets
containing less than 5,000 reviews or cases, and completing with datasets including over 100,000 comments.
According to the data depicted in Table 6, it is evident that the dataset group with a size of less than
20,000 reviews or comments exhibits the majority of literature which is 42.86%. This observation implies
that a dataset of this magnitude is enough to achieve effective training and generalization outcomes.
Approximately 14% of the literature surveyed used datasets above a size of 100,000. The observed outcome
can be explained by the fact that certain machine-learning methods require a smaller dataset. Therefore, it is
feasible to conduct a comprehensive examination of the performance of ML and DL algorithms based on the
varying sizes of datasets.
RQ3: what are the most common performance evaluation metrics used? The purpose of this part of
the findings is to provide the answer to the third question, which focuses on an analysis of the evaluation
metric. This research question also aims to limit down potential methods by providing a comparison of the
performance evaluation statistic. Validating the effectiveness of a text classification model can be done with
the use of a wide variety of performance evaluation criteria. These metrics can also be used to determine how
accurate a learning model such as machine and DL was trained.

Table 6. List of literature and dataset size

Literature Dataset size
[23], [25], [29], [31], [44], [45] <5,000
[20], [22], [26], [30], [43], [46] 5,000-20,000
[24], [36], [38], [47] 20,000-50,000
[21] 50,000-100,000
[33], [34], [35], [40] >100,000
[27], [28], [32], [37], [39], [41], [42] Not available

When assessing the efficacy of a ML model, accuracy is a crucial metric to take into consideration.
The model has successfully learned the relationships among the input samples if the output value is very
accurate, and it is then ready to classify future values. The most common metric for evaluating classification
performance is accuracy. However, it is important to note that accuracy may not provide an accurate
representation of performance when dealing with imbalanced datasets [48]. The F1 measure is recommended
for evaluating skewed datasets [49]. If DL models are employed and a decline in precision is observed, an
alternative measure such as macro-F1 can be utilized [50]. According to the data presented in Table 7, it is
observed that a significant proportion of the literature reviewed uses accuracy as the primary performance
criterion for evaluating their respective models. The implementation of machine and DL approaches is
common in the field. It is noticeable that few literatures used Perception score, mapping quality
characteristics, and exploring the relationship between internet reviews and hotel sales.
RQ4: what are the techniques/approaches gives higher accuracy? The evaluation of a text
classification model’s performance heavily relies on the aspect of accuracy, which is considered to be of
utmost importance. A value that exhibits a high level of accuracy signifies that the model has effectively
acquired a comprehensive understanding of the connections between the input samples.

Text classification on user feedback: a systematic literatures review (Zuleaizal Sidek)

1264  ISSN: 2502-4752

Table 7. List of literature and its evaluation metric

Literature Performance metric
[21], [25], [41], [47] Accuracy
[28], [32] Accuracy, precision, recall
[30], [44] Accuracy, F1-score
[20], [23], [24], [26], [29], [31], [36], [37] Accuracy, precision, recall, F1-score
[22], [27] Precision, recall, F1-score
[38] F1-score
[34] Perception score
[46] Mapping quality characteristic
[40] Relationship (online reviews–hotel sales)
[43] Classify user component
[33] Confusion matrix

Consequently, it is deemed proficient at classifying forthcoming values. According to the data

presented in Table 8, it can be observed that the random forest technique exhibits the highest level of
accuracy among the ML algorithms, achieving a remarkable accuracy rate of 99.2% when applied to the
TripAdvisor dataset. The following technique involved is the utilization of bidirectional encoder
representations from transformers (BERT) and support vector machine (SVM) techniques, which achieved
the highest accuracies of 98.8% and 97.17% respectively. It was observed that the SVM is the predominant
ML model employed for text classification, as evidenced by its appearance in 9 research papers. The SVM
achieved a peak accuracy of 97.17% when applied to the e-commerce dataset. The SVM is a powerful
classifier capable of achieving high performance in various circumstances, particularly in multilabel
problems. It exhibits the ability to handle both linear and nonlinear classification tasks [51]. The model’s
long-short term memory (LSTM) is also widely recognized and utilized in the field of text classification as it
produces accuracy as high as 97%. The data presented in Table 8 also demonstrates that most of the ML
algorithms achieved accuracy levels exceeding 90%.

Table 8. List of top ten accuracy score

Literature Dataset Technique Accuracy
[29] Tripadvisor Random forest 99.20%
[30] ATIS and Snips BERT 98.8%
[36] Mobile banking SVM 97.17%
[26] Twitter LSTM 97%
[23] E-commerce LSTM 96.92%
[44] E-commerce SVM 96.57%
[33] Web news SVM 95.04%
[31] ATIS SVM 95%
[24] Amazon Random forest 94%
[47] Amazon CNN 92.30%

4. CONCLUSION
This paper provides a comprehensive review of the distribution of publications on text classification
across several domains. This research presents the contributions by providing a critical analysis of the text
classification methodologies used in recent studies, along with the evaluation metrics used to assess the
model’s performance. A systematic literature review was implemented using the PRISMA as methodology,
which consists of a selection procedure, identification, screening, and eligibility. From the original pool of
484 retrieved materials, 28 literatures were selected after a thorough assessment based on four research
questions. The academic publications chosen for this study’s systematic reviews were published between
2019 and 2023 by four respectable publishers. Consists in detail: the type of datasets used, the size of
datasets, the classification type, the evaluation metrics and the best performance in terms of accuracy. SVM
is the main text classification method, delivering greater accuracy. For text classification, numerous datasets
from multiple areas were examined. The most commonly used datasets in literature published in the last five
years were Amazon customer reviews and tweets from Twitter. The popularity of these datasets is due to the
ease of crawling and extracting these data. Based on our comprehensive review of the articles, it is evident
that the adoption of machine learning and DL methodologies holds significant potential for advancing the
field of text classification. By using these techniques, organizations may effectively streamline sentiment
analysis processes, resulting in notable reductions in both cost and time spending. As for future work, the
utilization of machine learning and DL techniques could improve efficacy and minimize dependency on
human judgment in classifying consumer feedback and reviews.

Indonesian J Elec Eng & Comp Sci, Vol. 34, No. 2, May 2024: 1258-1267
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1265

ACKNOWLEDGEMENTS
The authors express their sincere appreciation to the reviewers for their valuable feedback, which
helped enhance the quality of this article. The funding for this research was provided by ‘Skim Zamalah
UTeM’ of Universiti Teknikal Malaysia Melaka (UTeM), Malaysia.

REFERENCES
[1] N. Krishnadoss and L. K. Ramasamy, “A study on high dimensional big data using predictive data analytics model,” Indonesian
Journal of Electrical Engineering and Computer Science (IJEECS), vol. 30, no. 1, pp. 174–182, Apr. 2023,
doi: 10.11591/ijeecs.v30.i1.pp174-182.
[2] A. K. Al-Mashhadany, M. B. Mohammed, and M. M. Alrawi, “Inheritance issues’ features extraction using Arabic text analyzer
(IFAA),” Indonesian Journal of Electrical Engineering and Computer Science (IJEECS), vol. 28, no. 1, pp. 611–624, Oct. 2022,
doi: 10.11591/ijeecs.v28.i1.pp611-624.
[3] R. A. Albeer, H. F. Al-Shahad, H. J. Aleqabie, and N. D. Al-Shakarchy, “Automatic summarization of YouTube video
transcription text using term frequency-inverse document frequency,” Indonesian Journal of Electrical Engineering and
Computer Science (IJEECS), vol. 26, no. 3, pp. 1512–1519, Jun. 2022, doi: 10.11591/ijeecs.v26.i3.pp1512-1519.
[4] A. Ibrahim, D. R. Indah, and D. I. Meytri, “The implementation of social customer relationship management for tourism
information system,” Indonesian Journal of Electrical Engineering and Computer Science (IJEECS), vol. 24, no. 3,
pp. 1578–1588, Dec. 2021, doi: 10.11591/ijeecs.v24.i3.pp1578-1588.
[5] M. Mkono and J. Tribe, “Beyond reviewing: uncovering the multiple roles of tourism social media users,” Journal of Travel
Research, vol. 56, no. 3, pp. 287–298, Aug. 2017, doi: 10.1177/0047287516636236.
[6] J. Joung, K. Jung, S. Ko, and K. Kim, “Customer complaints analysis using text mining and outcome-driven innovation method for
market-oriented product development,” Sustainability (Switzerland), vol. 11, no. 1, p. 40, Dec. 2019, doi: 10.3390/su11010040.
[7] P. Zhou and N. El-Gohary, “Ontology-based multilabel text classification of construction regulatory documents,” Journal of
Computing in Civil Engineering, vol. 30, no. 4, Jul. 2016, doi: 10.1061/(asce)cp.1943-5487.0000530.
[8] V. Korde, “Text classification and classifiers:a survey,” International Journal of Artificial Intelligence & Applications, vol. 3,
no. 2, pp. 85–99, Mar. 2012, doi: 10.5121/ijaia.2012.3208.
[9] P. Taylor, “Volume of data/information created, captured, copied, and consumed worldwide from 2010 to 2020, with forecasts
from 2021 to 2025,” Statista, 2023. https://www.statista.com/statistics/871513/worldwide-data-created/ (accessed Sep. 16, 2023).
[10] V. Gupta and G. S. Lehal, “A survey of text mining techniques and applications,” Journal of Emerging Technologies in Web
Intelligence, vol. 1, no. 1, pp. 60–76, Aug. 2009, doi: 10.4304/jetwi.1.1.60-76.
[11] M. A. Tayal, V. Bajaj, A. Gore, P. Yadav, and V. Chouhan, “Automatic domain classification of text using machine learning,” in
2023 International Conference on Communication, Circuits, and Systems, IC3S 2023, May 2023, pp. 1–5,
doi: 10.1109/IC3S57698.2023.10169470.
[12] R. Risnantoyo, A. Nugroho, and K. Mandara, “Sentiment analysis on corona virus pandemic using machine learning algorithm,”
Journal of Informatics and Telecommunication Engineering, vol. 4, no. 1, pp. 86–96, Jul. 2020, doi: 10.31289/jite.v4i1.3798.
[13] V. S. Anoop and S. Asharaf, “Aspect-oriented sentiment analysis: a topic modeling-powered approach,” Journal of Intelligent
Systems, vol. 29, no. 1, pp. 1166–1178, Dec. 2020, doi: 10.1515/jisys-2018-0299.
[14] L. M. Rojas-Barahona, “Deep learning for sentiment analysis,” Language and Linguistics Compass, vol. 10, no. 12, pp. 701–719,
Dec. 2016, doi: 10.1111/lnc3.12228.
[15] B. Liu, “Sentiment analysis and opinion mining,” Synthesis Lectures on Human Language Technologies, vol. 5, no. 1, pp. 1–167,
May 2012, doi: 10.2200/S00416ED1V01Y201204HLT016.
[16] A. T. Al-Tuama and D. A. Nasrawi, “Intent classification using machine learning algorithms and augmented data,” in 2022
International Conference on Data Science and Intelligent Computing, ICDSIC 2022, Nov. 2022, pp. 234–239,
doi: 10.1109/ICDSIC56987.2022.10075794.
[17] H. Zhang, M. A. Babar, and P. Tell, “Identifying relevant studies in software engineering,” Information and Software Technology,
vol. 53, no. 6, pp. 625–637, Jun. 2011, doi: 10.1016/j.infsof.2010.12.010.
[18] H. Snyder, “Literature review as a research methodology: an overview and guidelines,” Journal of Business Research, vol. 104,
pp. 333–339, Nov. 2019, doi: 10.1016/j.jbusres.2019.07.039.
[19] O. Kaiwartya et al., “Internet of vehicles: motivation, layered architecture, network model, challenges, and future aspects,”
IEEE Access, vol. 4, pp. 5356–5373, 2016, doi: 10.1109/ACCESS.2016.2603219.
[20] R. Akila and S. Revathi, “Opinion mining of social media reviews using fuzzy inference and intent prediction using deep
learning,” Advances in Engineering Software, vol. 174, p. 103293, Dec. 2022, doi: 10.1016/j.advengsoft.2022.103293.
[21] A. M. Alayba and V. Palade, “Leveraging Arabic sentiment classification using an enhanced CNN-LSTM approach and effective
Arabic text preparation,” Journal of King Saud University - Computer and Information Sciences, vol. 34, no. 10, pp. 9710–9722,
Nov. 2022, doi: 10.1016/j.jksuci.2021.12.004.
[22] W. Huang, M. Lin, and Y. Wang, “Sentiment analysis of Chinese E-commerce product reviews using ERNIE word embedding
and attention mechanism,” Applied Sciences (Switzerland), vol. 12, no. 14, p. 7182, Jul. 2022, doi: 10.3390/app12147182.
[23] A. Sharma and M. O. Shafiq, “A comprehensive artificial intelligence based user intention assessment model from online reviews
and social media,” Applied Artificial Intelligence, vol. 36, no. 1, Jan. 2022, doi: 10.1080/08839514.2021.2014193.
[24] Prathyakshini, K. P. Salian, B. R. Puneeth, T. Nargis, and S. Salian, “A novel approach for classification of online product
reviews using various machine learning techniques,” in 6th International Conference on Electronics, Communication and
Aerospace Technology, ICECA 2022 - Proceedings, Dec. 2022, pp. 878–884, doi: 10.1109/ICECA55336.2022.10009447.
[25] Z. A. Diekson, M. R. B. Prakoso, M. S. Q. Putra, M. S. A. F. Syaputra, S. Achmad, and R. Sutoyo, “Sentiment analysis for customer
review: case study of Traveloka,” Procedia Computer Science, vol. 216, pp. 682–690, 2022, doi: 10.1016/j.procs.2022.12.184.
[26] W. Aljedaani et al., “Sentiment analysis on Twitter data integrating TextBlob and deep learning models: the case of US airline
industry,” Knowledge-Based Systems, vol. 255, p. 109780, Nov. 2022, doi: 10.1016/j.knosys.2022.109780.
[27] M. Tummawat and C. Jareanpon, “The sentiment classification of hotel reviews and hotel description using feature-based
technique for customer relationship management,” ICIC Express Letters, vol. 17, no. 3, pp. 279–287, 2023,
doi: 10.24507/icicel.17.03.279.
[28] L. Gunawan, M. S. Anggreainy, L. Wihan, Santy, G. Y. Lesmana, and S. Yusuf, “Support vector machine based emotional
analysis of restaurant reviews,” Procedia Computer Science, vol. 216, pp. 479–484, 2022, doi: 10.1016/j.procs.2022.12.160.

Text classification on user feedback: a systematic literatures review (Zuleaizal Sidek)

1266  ISSN: 2502-4752

[29] C. Huda, Y. Heryadi, Lukas, and W. Budiharto, “Aspect-based sentiment analysis in tourism industry for tourism recommender
system,” in 2022 5th International Seminar on Research of Information Technology and Intelligent Systems, ISRITI 2022,
Dec. 2022, pp. 407–412, doi: 10.1109/ISRITI56927.2022.10053083.
[30] Y. Chen and Z. Luo, “Pre-trained joint model for intent classification and slot filling with semantic feature fusion,” Sensors,
vol. 23, no. 5, p. 2848, Mar. 2023, doi: 10.3390/s23052848.
[31] M. Menda and G. S. Keerthi, “Intent classification in conversational system using machine learning techniques,” International
Journal of Computer Applications, vol. 183, no. 51, pp. 6–11, Feb. 2022, doi: 10.5120/ijca2022921913.
[32] A. Ghazzawi and B. Alharbi, “Analysis of customer complaints data using data mining techniques,” Procedia Computer Science,
vol. 163, pp. 62–69, 2019, doi: 10.1016/j.procs.2019.12.087.
[33] A. Mulahuwaish, K. Gyorick, K. Z. Ghafoor, H. S. Maghdid, and D. B. Rawat, “Efficient classification model of web news
documents using machine learning algorithms for accurate information,” Computers and Security, vol. 98, p. 102006, Nov. 2020,
doi: 10.1016/j.cose.2020.102006.
[34] S. Dhar and I. Bose, “Walking on air or hopping mad? Understanding the impact of emotions, sentiments and reactions on ratings
in online customer reviews of mobile apps,” Decision Support Systems, vol. 162, p. 113769, Nov. 2022,
doi: 10.1016/j.dss.2022.113769.
[35] S. Park, J. Cho, K. Park, and H. Shin, “Customer sentiment analysis with more sensibility,” Engineering Applications of Artificial
Intelligence, vol. 104, p. 104356, Sep. 2021, doi: 10.1016/j.engappai.2021.104356.
[36] N. A. Rahman, S. D. Idrus, and N. L. Adam, “Classification of customer feedbacks using sentiment analysis towards mobile
banking applications,” IAES International Journal of Artificial Intelligence, vol. 11, no. 4, pp. 1579–1587, Dec. 2022,
doi: 10.11591/ijai.v11.i4.pp1579-1587.
[37] A. Londhe and P. V. R. D. P. Rao, “Dynamic classification of sentiments from restaurant reviews using novel fuzzy-encoded
LSTM,” International Journal on Recent and Innovation Trends in Computing and Communication, vol. 10, no. 9, pp. 112–124,
Sep. 2022, doi: 10.17762/ijritcc.v10i9.5714.
[38] Á. Aldunate, S. Maldonado, C. Vairetti, and G. Armelini, “Understanding customer satisfaction via deep learning and natural
language processing,” Expert Systems with Applications, vol. 209, p. 118309, Dec. 2022, doi: 10.1016/j.eswa.2022.118309.
[39] A. Patel, P. Oza, and S. Agrawal, “Sentiment analysis of customer feedback and reviews for airline services using language
representation model,” Procedia Computer Science, vol. 218, pp. 2459–2467, 2022, doi: 10.1016/j.procs.2023.01.221.
[40] Y. Sim, S. K. Lee, and I. Sutherland, “The impact of latent topic valence of online reviews on purchase intention for the
accommodation industry,” Tourism Management Perspectives, vol. 40, p. 100903, Oct. 2021, doi: 10.1016/j.tmp.2021.100903.
[41] V. Desai, S. Wadhwa, A. A, and B. Bajaj, “Text-based intent analysis using deep learning,” International Journal of Innovative
Science and Research Technology, vol. 5, no. 7, pp. 267–274, Jul. 2020, doi: 10.38124/ijisrt20jul342.
[42] V. R. Azhaguramyaa, J. Janet, G. R. Madhavan, S. Balakrishnan, and K. Arunkumar, “Sentiment analysis on book reviews using
machine learning techniques,” in 8th International Conference on Advanced Computing and Communication Systems, ICACCS
2022, Mar. 2022, pp. 1530–1534, doi: 10.1109/ICACCS54159.2022.9785311.
[43] Y. Son and W. Kim, “Development of methodology for classification of user experience (UX) in online customer review,”
Journal of Retailing and Consumer Services, vol. 71, p. 103210, Mar. 2023, doi: 10.1016/j.jretconser.2022.103210.
[44] P. Vijayaragavan, R. Ponnusamy, and M. Aramudhan, “An optimal support vector machine based classification model for
sentimental analysis of online product reviews,” Future Generation Computer Systems, vol. 111, pp. 234–240, Oct. 2020,
doi: 10.1016/j.future.2020.04.046.
[45] P. Yu, X. Zhou, Z. Li, and Y. Yan, “Inactivation and change of tetracycline-resistant Escherichia coli in secondary effluent by
visible light-driven photocatalytic process using Ag/AgBr/g-C3N4,” Science of The Total Environment, vol. 705, p. 135639,
Feb. 2020, doi: 10.1016/j.scitotenv.2019.135639.
[46] H. Liu, S. Wu, C. Zhong, and Y. Liu, “The effects of customer online reviews on sales performance: the role of mobile phone’s
quality characteristics,” Electronic Commerce Research and Applications, vol. 57, 2023, doi: 10.1016/j.elerap.2022.101229.
[47] L. Arora, P. Srivastava, and S. A. Kumar, “Evaluation of product reviews using deep learning classifier models,” in 2022 13th
International Conference on Computing Communication and Networking Technologies, ICCCNT 2022, Oct. 2022, pp. 1–5,
doi: 10.1109/ICCCNT54827.2022.9984463.
[48] B. Espejo-Garcia, J. Martinez-Guanter, M. Pérez-Ruiz, F. J. Lopez-Pellicer, and F. J. Zarazaga-Soria, “Machine learning for
automatic rule classification of agricultural regulations: a case study in Spain,” Computers and Electronics in Agriculture,
vol. 150, pp. 343–352, Jul. 2018, doi: 10.1016/j.compag.2018.05.007.
[49] A. Ligthart, C. Catal, and B. Tekinerdogan, “Analyzing the effectiveness of semi-supervised learning approaches for opinion
spam classification,” Applied Soft Computing, vol. 101, p. 107023, Mar. 2021, doi: 10.1016/j.asoc.2020.107023.
[50] D. Song, A. Vold, K. Madan, and F. Schilder, “Multi-label legal document classification: A deep learning-based approach with label-
attention and domain-specific pre-training,” Information Systems, vol. 106, p. 101718, May 2022, doi: 10.1016/j.is.2021.101718.
[51] N. A. P. Rostam and N. H. A. H. Malim, “Text categorisation in Quran and Hadith: overcoming the interrelation challenges using
machine learning and term weighting,” Journal of King Saud University - Computer and Information Sciences, vol. 33, no. 6,
pp. 658–667, Jul. 2021, doi: 10.1016/j.jksuci.2019.03.007.

BIOGRAPHIES OF AUTHORS

Zuleaizal Sidek is the founder and managing director of Kencana Niaga, an IT

start-up and independent provider of data analytics for interdisciplinary analysis. He is also a
postgraduate student in Philosophy Doctor in Data Science at the Universiti Teknikal Malaysia
(UTeM). He is currently working as Data Scientist at the Institut Tun Perak (Melaka state
owned company). He has multiple experiences in IT for more than 10 years at the Universiti
Teknologi MARA as an IT Manager. His first paper on e-government services: a value of
E-service in local government: a fuzzy approach evaluation”. His areas of interest are big data
management and analytics, machine learning and blockchain technology. He can be contacted
at email: [email protected].

Indonesian J Elec Eng & Comp Sci, Vol. 34, No. 2, May 2024: 1258-1267
Indonesian J Elec Eng & Comp Sci ISSN: 2502-4752  1267

Assoc. Prof. Dr. Sharifah Sakinah Syed Ahmad is currently an associate

professor in the Department of Intelligent Computing and Analytics (ICA), Faculty of
Information and Communication Technology, Universiti Teknikal Malaysia Melaka (UTeM).
She received her bachelor’s and master’s degrees in applied mathematics from the School of
Mathematics at the University of Science, Malaysia. Following this, she received her Ph.D.
from the University of Alberta, Canada in 2012 in intelligent systems. She can be contacted at
email: [email protected].

Dr. Yogan Jaya Kumar is a Senior Lecturer at the Department of Intelligent

Computing and Analytic in the Faculty of Information and Communication Technology
(FTMK), Universiti Teknikal Malaysia Melaka (UTeM). He earned both his bachelor degree
and master degree from Universiti Sains Malaysia (USM), in the field of Mathematical
Science in year 2003 and 2005. He joined the Centre for Affiliate and Diploma Programme,
MMU as assistant lecturer in 2006. He was then offered a lecturer position at UTeM in 2008.
He went on to complete his Ph.D. studies a Universiti Teknologi Malaysia, in 2014 in the field
of Computer Science under the supervision of Professor Dr. Naomie Salim. Currently, his
research involves in the field of text mining, information extraction and AI applications.
He has published a number of journal articles in applied soft computing, mainly in the field of
text document summarization. He can be contacted at email: [email protected].

Dr. Noor Hasimah Ibrahim Teo is a senior lecturer at the Universiti Teknologi
Teknologi MARA Melaka, Malaysia. She received her Ph.D. in Computer Science from the
University of Warwick, United Kingdom in the year 2019 and an M.Sc. in Computer Science
from Universiti Teknologi MARA, Malaysia. She received Ernst Mach Grant ASEA-
UNINET, OeAD Austria for a postdoctoral visit to the Technological University of Vienna in
2022. Her research interests are concentrated in the field of ontology, text processing,
machine learning and question answering system. She can be contacted at email:
[email protected].

Text classification on user feedback: a systematic literatures review (Zuleaizal Sidek)