Sentiment analysis with machine learning and deep learning: A survey of techniques

and applications
Nikhil Sanjay Suryawanshi *

California, USA.

International Journal of Science and Research Archive, 2024, 12(02), 005–015

Publication history: Received on 19 May 2024; revised on 26 June 2024; accepted on 29 June 2024

Article DOI: https://doi.org/10.30574/ijsra.2024.12.2.1205

Abstract
Sentiment analysis is the task of automatically identifying the sentiment expressed in text. It has become increasingly
important in many applications such as social media monitoring, product reviews analysis, and customer feedback
evaluation. With the advent of deep learning techniques, sentiment analysis has seen significant improvements in
performance and accuracy. This paper presents a comprehensive survey of machine learning and deep learning
methods for sentiment analysis at the document, sentence, and aspect levels. We first provide an overview of traditional
machine learning approaches to sentiment analysis and their limitations. We then look into various machine learning
and deep learning architectures that have been successfully applied to this task. Additionally, we discuss the challenges
of dealing with different data modalities, such as visual and multimodal data, and how both techniques have been
adapted to address these challenges. Furthermore, we explore the applications of sentiment analysis in diverse domains,
including social media, product reviews, and healthcare. Finally, we highlight the current limitations of deep learning
approaches for sentiment analysis and outline potential future research directions. This survey aims to provide
researchers and practitioners with a comprehensive understanding of the state-of-the-art deep learning techniques for
sentiment analysis and their practical applications.

Keywords: Natural Language Processing; Sentiment Analysis; Text Analysis; Recurrent Neural Network; Deep Neural
Network; Convolutional Neural Network; Machine Learning; Deep Learning

1. Introduction
Sentiment analysis, a branch of natural language processing, focuses on the automated identification and classification
of emotions, opinions, and subjective expressions conveyed through written text [1]. With the exponential rise of social
media platforms, the abundance of publicly shared sentiments and viewpoints has increased significantly, rendering
sentiment analysis an indispensable tool for gauging public opinion across various domains, including business, politics,
and others. This technique involves analyzing diverse sources, such as tweets, reviews, and other forms of user-
generated content, to extract valuable insights into prevalent attitudes and perceptions.

Sentiment analysis has proven to be an effective predictor of significant events, such as a movie's box office performance
and election outcomes [2]. The process entails evaluating opinions expressed about specific entities, like products,
services, individuals, or locations, which can be found on various websites and platforms like Amazon and Flipkart.
These opinions are typically categorized as positive, negative, or neutral. The primary objective of sentiment analysis is
to automatically determine the sentiment conveyed in user-generated content [3], enabling organizations and decision-
makers to understand and respond to public sentiment effectively.

* Corresponding author: Nikhil Sanjay Suryawanshi

Copyright © 2024 Author(s) retain the copyright of this article. This article is published under the terms of the Creative Commons Attribution Liscense 4.0.
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

Moreover, the ever-increasing volume of unstructured data generated on social media platforms has fueled the demand
for sentiment analysis techniques capable of analyzing and structuring this vast trove of information [4]. By harnessing
sentiment analysis, businesses and organizations can gain a competitive edge by tailoring their strategies and decision-
making processes to align with public sentiment, ultimately enhancing customer satisfaction and fostering stronger
relationships with their target audiences.

Traditional machine learning techniques have been widely employed for sentiment analysis tasks, involving careful
feature engineering and the use of algorithms such as Naive Bayes, Support Vector Machines, and Logistic Regression.
However, these methods often struggle to capture the intricacies of natural language and require extensive manual
effort in feature extraction. In recent years, deep learning approaches have emerged as powerful alternatives, leveraging
their ability to automatically learn discriminative features from raw text data. Architectures like Recurrent Neural
Networks (RNNs), Long Short-Term Memory (LSTM) networks, and Convolutional Neural Networks (CNNs) have
shown remarkable success in sentiment analysis tasks, effectively modeling the sequential nature of text and capturing
long-range dependencies. Furthermore, the advent of pre-trained language models, such as BERT and GPT, has
revolutionized the field, enabling transfer learning and achieving state-of-the-art performance on sentiment analysis
tasks with relatively smaller datasets. These deep learning techniques have significantly advanced the capabilities of
sentiment analysis systems, enabling more accurate and robust analysis of sentiment across various domains and
applications. Figure 2 provides a visual representation of the sentiment analysis process.

Figure 1 Sentiment analysis approaches can be categorized as machine learning and deep learning.

Figure 2 Flow charts of the sentiment analysis process using conventional machine learning, and deep learning
approaches.

In this comprehensive survey, we provide an in-depth exploration of both traditional machine learning and modern
deep learning approaches for sentiment analysis tasks. We discuss conventional machine learning techniques like Naive
Bayes, Support Vector Machines, and Logistic Regression, as well as deep learning architectures such as recurrent neural
networks (RNNs), long short-term memory (LSTM) networks, convolutional neural networks (CNNs), and attention-
based models. We examine the challenges of multimodal sentiment analysis and how deep learning models address

6
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

them. Additionally, we explore the practical applications of sentiment analysis across domains like social media, product
reviews, and healthcare. Finally, we outline the current limitations and future research directions, including handling
context, sarcasm, domain adaptation, interpretability, and model robustness in machine learning and deep learning for
sentiment analysis.

The rest of the paper is organized as section II provides a literature survey of state of art systems for sentiment analysis
using machine learning and deep learning models. Section III provides a discussion and section IV concludes the paper
with future research directions.

2. Literature Survey
Machine learning and deep learning are two areas of artificial intelligence that have become very popular lately. Machine
learning is a way of analyzing data that helps computers build models to make predictions or decisions without being
manually programmed. Deep learning, which is a part of machine learning, uses neural networks to handle more
complicated tasks. These models are inspired by how the human brain works and can learn from data that isn't neatly
organized or labeled[5-11].

There has been a lot of research and progress in machine learning and deep learning, especially as computers have
gotten faster and more data has become available. One of the biggest benefits of these technologies is that they can learn
from data on their own, without needing to be explicitly programmed. This makes them very useful for tasks like
recognizing images and speech, understanding natural language, and making recommendations[12-19].

In recent years, the increase in data and better computing power has led to more advanced algorithms and models in
machine learning and deep learning. This has resulted in significant breakthroughs in many industries, showing just
how powerful these technologies can be in solving real-world problems[20-24].

2.1. Machine learning approaches

Researchers in [25] utilized a cut-based approach for subjectivity classification on a movie review dataset. The cut-
based subjectivity detectors identified subjective portions of documents by examining pair-wise relationships and per-
item information. Initially, they extracted the subjective sections from the review documents and classified them into
subjective and objective categories using a minimum-cut framework. The experiments demonstrated that employing a
minimum-cut framework enhanced the effectiveness of sentiment analysis.

In a comparative experiment detailed in [26], different machine learning classifiers were tested on a product reviews
dataset. The study employed both a generative model and a discriminative algorithm for classification, specifically
testing the winnow algorithm [27]. The findings indicated that Naive Bayes (NB) outperformed Support Vector
Machines (SVM) when using unigram features on datasets other than movie reviews. The results also revealed that high-
order n-grams could help differentiate article polarity in mixed contexts. However, the authors noted that a large
number of n-grams could introduce noise, suggesting the need for a better feature selection method.

In [28], the authors addressed a sentiment analysis model in the medical field using an SVM classifier and a language
model. Instead of identifying personal opinions, they extracted polarity information from medical outcomes in patient
records. A small lexicon of medical domain words was manually compiled. By incorporating domain knowledge and
linguistic features, the model's performance was enhanced.

The paper in [29] compared SVM and Naive Bayes sentiment classifiers on a Cantonese review dataset. Various n-gram
representations, including unigram, bigram, and trigram with frequency features, were used to study the impact of
feature size and representation on classification performance. The results showed that the NB classifier achieved better
performance than SVM, with a 95.67% accuracy for 900-1100 features. Similarly, researchers in [30] compared lexicon-
based, machine learning, and deep learning methods for sentiment analysis on Manipuri video review comments. Their
manually categorized dataset of 4000 comments, collected from social media, revealed that the NB classifier
outperformed others. Another study [31] introduced a sentiment analysis system for Manipuri news using a machine
learning approach. Data were gathered from local newspapers, and language-specific preprocessing tasks, such as
transliteration, building a negative morpheme-based lexicon, and filtering out noisy words, were performed. These
efforts improved the system's robustness and achieved the highest accuracy.

In 2019, a novel approach to sentiment analysis was introduced by combining boosting and bagging methods within an
AdaBoost model[32]. This study focused on analyzing US airline Twitter data, which was preprocessed to remove

7
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

irrelevant information and then analyzed using data-mining techniques. A 75-25 data split was used, with 75% for
training and 25% for testing, resulting in an F-score of 68% for the AdaBoost model.

Another 2019 study applied the ridge classifier method to sentiment analysis on the IMDb and Sentiment140
datasets[33]. The preprocessing stage included tokenization, case folding, and the removal of URLs, HTML tags, and
irrelevant words, with the data represented using the TF-IDF method. The ridge classifier achieved 90.54% accuracy on
the IMDb dataset and 76.84% on the Sentiment140 dataset, demonstrating its effectiveness.

A study in 2020 aimed to improve sentiment analysis accuracy for Arabic Tweets using the multinomial naive Bayes
method on a dataset of 2000 labeled Tweets[34]. Preprocessing involved tokenization using 4-grams and stemming
with the Khoja stemmer, with the data represented as TF-IDF features and classified using a fivefold cross-validation
process, achieving an accuracy of 87.5%.

In a comprehensive 2020 study, sentiment analysis of US airline Twitter data was conducted using six different machine
learning models: support vector machine (SVM), logistic regression, random forest, XgBoost, naive Bayes, and decision
tree[35]. The data underwent preprocessing including stop-word removal, punctuation removal, case folding, and
stemming. Feature extraction was performed using the bag-of-words method on a dataset from CrowdFlower and
Kaggle, containing 14,640 samples with three sentiment classes: positive, negative, and neutral. SVM achieved the
highest accuracy of 83.31%, followed by logistic regression at 81.81%, using a 70% training and 30% testing split.

In 2021, a study on sentiment analysis related to violence against women utilized several machine-learning models,
including SVM, k-nearest neighbor, naive Bayes, and decision tree[36]. The dataset, collected from Arabic Tweets, was
preprocessed with tokenization, stemming, and stop-word removal, and feature extraction was done using the TF-IDF
method. SVM achieved the highest accuracy of 78.25%.

Another 2021 study developed a sentiment analyzer to classify text polarity with high accuracy using naive Bayes,
multinomial naive Bayes, Bernoulli naive Bayes, logistic regression, and linear support vector classification
methods[37]. The 'Twitter samples' corpus from the Natural Language Toolkit, consisting of 10,000 Tweets with equal
numbers of positive and negative Tweets, was used. Preprocessing included tokenization, stop-word removal, URL
removal, symbol removal, case folding, and lemmatization. The naive Bayes method achieved the highest accuracy of
99.73%, making it the best-performing algorithm in the study.

2.2. Deep Learning Approaches

The deep learning-based sentiment analysis model has gained popularity due to its exceptional performance. Deep
learning, an emerging area within machine learning, offers both supervised and unsupervised feature representation
approaches[38]. This technique utilizes multiple perceptron layers, inspired by the human brain[39]. A foundational
contribution to deep learning was made in 2006 with the introduction of a deep belief network[40]. One notable deep
learning model, the Recursive Neural Tensor Network (RNTN), was developed for sentence-level sentiment
classification by modeling the composition of phrases. This model was used alongside the Stanford Sentiment Treebank
corpus, which includes parse trees annotated with sentiment[41]. Fine-tuning a pre-trained network often yields better
performance compared to using a feature vector alone, as it better captures discriminative information[42].

In one study, researchers adopted a convolutional neural network (CNN) architecture for textual sentiment
classification[43]. Their system combined an embedding layer, two convolutional layers, a pooling layer, and a fully
connected layer. Using datasets of customer reviews, movie reviews, and the Stanford Sentiment Treebank, they
compared traditional machine learning classifiers to their deep neural architecture. Their findings showed that the
parallel CNN architecture reduced computational overhead and improved classification accuracy, particularly with long
texts.

Researchers continue to integrate various deep-learning techniques for sentiment classification. One study combined
different deep learning architectures into a single framework, using pre-trained word vectors as input along with a
convolutional neural network (CNN) and long short-term memory (LSTM) layers[44]. This combination leverages CNN
for local text features and LSTM for context-dependent features, resulting in superior performance compared to existing
LSTM, CNN, CNN-LSTM, and SVM classifiers.

Another study reported a document-level sentiment analysis framework using a deep CNN and bidirectional LSTM
(BiLSTM) on French news articles[45]. The CNN extracted local features, which were then fed into a BiLSTM. After
applying multiple filters and max pooling, the concatenated results were input into a fully connected layer, with a

8
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

softmax function used for classification. This method achieved high performance in sequence classification tasks on long
texts.

In research focused on code-mixed languages, a convolutional neural network was applied to a Bengali-English dataset,
as well as a monolingual Telugu dataset[46]. The code-mixed Bengali data achieved higher accuracy compared to the
Telugu dataset, likely due to the morphological richness of Telugu, which increases the likelihood of words having the
same meaning. Another framework utilized CNN and LSTM-based ensemble models to detect Hinglish hate speech on
social media, reporting an F-Score of 0.617 on the test data[47].

Recent studies have begun incorporating attention mechanisms to enhance performance. A hierarchical attention
mechanism was used in a document-level sentiment classification model to focus on the essential content of documents,
resulting in improved performance on text-based review datasets[48]. Another study addressed the challenge of
generating semantic relationships between sentences in a document by developing the SR-LSTM model, which uses
LSTM layers to extract sentence vectors and recognize inter-sentence relationships[49].

In a 2017 study, sentiment analysis was performed on a dataset of 4000 Tweets in both Korean and English. The
preprocessing phase included tokenization, case folding, stemming, and the removal of stop words, numbers, and
punctuation marks. A multilayer perceptron (MLP) model with three hidden layers, optimized using Stochastic Gradient
Descent (SGD), achieved an accuracy of 75.03% on the dataset [50].

Similarly, a 2019 study utilized an MLP model for sentiment analysis on Turkish Tweets containing the hashtag
"15Temmuz". The dataset consisted of 3000 positive and negative Tweets. The preprocessing phase included Turkish
deASCIIfication, tokenization, stop-word and punctuation removal, and stemming. The word2vec pre-trained model
was used to convert the text into embeddings, which served as inputs to the MLP with six dense layers and three dropout
layers, achieving an accuracy of 81.86% [51].

In 2021, an MLP was used for sentiment analysis on COVID-19-related Tweets, which included 101,435 Tweets. The
data were preprocessed to remove HTML tags and non-letters and to tokenize and stem the text. Text data were
represented using a count vectorizer and TF-IDF vectorizer, and classified using an MLP model with five hidden layers
and ReLU activation. The count vectorizer representation with MLP achieved the highest accuracy of 93.73% [52].

A 2018 study compared several machine learning and deep learning methods for sentiment analysis using the IMDb
dataset containing 3000 reviews. The text was preprocessed to remove irrelevant characters, symbols, repeating words,
and stop words, then represented using the count vectorizer. The CNN model achieved the highest accuracy of 99.33%
[53].

In 2020, machine learning and deep learning methods were compared for sentiment analysis using Sentiment140 and
Twitter US Airline Sentiment datasets. Preprocessing included tokenization, case folding, and the removal of stopwords,
URLs, hashtags, and punctuation. Methods compared included multinomial naive Bayes, logistic regression, SVM, LSTM,
and an ensemble of multinomial naive Bayes, logistic regression, and SVM with majority voting. LSTM recorded the
highest accuracy of 82% on Sentiment140, while SVM achieved 68.9% on the Twitter US Airline Sentiment dataset [54].

Sentiment analysis of 5000 Bangla Tweets using LSTM was conducted in 2019. The data were preprocessed by removing
spaces and punctuation marks and splitting the dataset into 80% training, 10% validation, and 10% testing. An LSTM
architecture with five layers, each with a size of 128, a batch size of 25, and a learning rate of 0.0001, achieved an
accuracy of 86.3% [55].

Another 2019 study analyzed sentiment in the Saudi dialect using LSTM and BiLSTM models. The dataset consisted of
60,000 Tweets with positive and negative sentiments. Preprocessing involved removing numbers, punctuation, special
symbols, and non-Arabic letters. Text was normalized and encoded using a word2vec pre-trained model for word
embedding. The dataset was split into 70% training and 30% testing, with BiLSTM achieving 94% accuracy [56]. A
recurrent neural filter-based CNN and LSTM model for sentiment analysis was proposed in 2018. The model utilized an
RNN as a convolutional filter and was trained and tested using the Stanford Sentiment Treebank dataset, encoded using
GloVe word embedding. The model included an embedding layer, pooling layer, and LSTM layer, using the Adam
optimizer and early stopping to prevent overfitting, achieving an accuracy of 53.4% [57].

A hybrid model combining CNNs and LSTM was proposed in 2019. The data comprised 32,000 Tweets from the
International Workshop on Semantic Evaluation competition. Preprocessing involved removing URLs, emoticons, and
special characters and converting text to lowercase. The hybrid model leveraged word2vec and GloVe embeddings, with

9
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

the CNN and LSTM models achieving 59% accuracy [58]. In 2020, a Bi-LSTM model was introduced for sentiment
analysis on Twitter data using the Twitter US Airline Sentiment dataset. Preprocessing included tokenization and the
removal of stopwords and punctuation. Text was represented using word2vec, GloVe, and sentiment-specific
embeddings, with the Bi-LSTM model achieving 81.20% accuracy [59].

A Bi-LSTM model for analyzing Amazon e-commerce reviews was proposed in 2021. The dataset contained 23,485
reviews divided into negative, neutral, and positive classes. Data were cleaned by tokenizing and removing special
characters, followed by word2vec embedding. The BiLSTM model achieved an accuracy of 90.26% [60]. Similarly, a
character-based deep bidirectional long short-term memory (DBLSTM) method was used for analyzing Tamil Tweets
in 2020. The dataset contained 1500 Tweets divided into positive, negative, and neutral classes. Preprocessing involved
removing unnecessary symbols and characters, with word2vec embedding used for DBLSTM, achieving 86.2% accuracy
[61].

A sentiment analysis bidirectional long short-term memory (SAB-LSTM) model was proposed in 2020. The model
included 196 Bi-LSTM units, 128 embedding layers, 4 dense layers, and a softmax activation classification layer. The
dataset of 80,689 samples from various sources was divided into 90% training and 10% testing, with SAB-LSTM
outperforming traditional LSTM models [62]. Lastly, a 2021 study explored the use of RNNs for sentiment analysis of
customer reviews from hotel booking websites. Preprocessing involved lemmatization, stemming, and the removal of
punctuation and stop-words. The LSTM and GRU models, with 30 and 25 hidden layers respectively, achieved accuracies
of 86% and 84% on the dataset [63].

Table 1 Comparative Analysis of state of art the of systems

Authors Dataset Preprocessing Model Accuracy

Techniques
[22] 4000 Tweets (Korean Tokenization, case folding, MLP with SGD (3 hidden 75.03%
& English) stemming, stop words, layers)
numbers, punctuation
removal
[23] 3000 Turkish Tweets Deasciification, MLP with word2vec (6 81.86%
(#15Temmuz) tokenization, stop-word and dense layers, 3 dropout
punctuation removal, layers)
stemming
[24] 101,435 COVID-19 HTML tags removal, MLP with ReLU 93.73%
Tweets tokenization, stemming, (5 hidden layers)
count vectorizer, TF-IDF
vectorizer
[25] IMDb dataset (3000 Removal of irrelevant CNN 99.33%
reviews) characters, symbols,
repeating words, stop
words, count vectorizer
[26] Sentiment140, Twitter Tokenization, case folding, LSTM, SVM, multinomial 82%
US Airline Sentiment stop-words, URLs, hashtags, naive Bayes, logistic (Sentiment140),
punctuation removal regression, ensemble 68.9%
model (US Airline)
[27] 5000 Bangla Tweets Space and punctuation LSTM 86.3%
removal, 80-10-10 train- (5 layers, 128 size)
validation-test split
[28] 60,000 Saudi Tweets Numbers, punctuation, LSTM, BiLSTM with 94%
special symbols, non-Arabic word2vec
letters removal,
normalization
[29] Stanford Sentiment GloVe word embedding RNN as a convolutional 53.4%
Treebank filter

10
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

[30] 32,000 Tweets URLs, emoticons, special Hybrid CNN and LSTM with 59%
(International characters removal, word2vec and GloVe
Workshop on Semantic lowercase conversion
Evaluation)
[38] Twitter US Airline Tokenization, stop-words, Bi-LSTM with word2vec, 81.20%
Sentiment punctuation removal GloVe, sentiment-specific
embeddings
[39] 23,485 Amazon Tokenization, special BiLSTM 90.26%
reviews characters removal,
word2vec embedding
[40] 1500 Tamil Tweets Unnecessary symbols, DBLSTM 86.2%
characters removal,
word2vec embedding
[41] 80,689 samples from - SAB-LSTM with 196 Bi- -
various sources LSTM units, 128
embedding layers, 4 dense
layers, softmax activation
[42] Customer reviews Lemmatization, stemming, LSTM, GRU (30 and 25 86% (LSTM), 84%
from hotel booking punctuation, stop-words hidden layers) (GRU)
websites removal

3. Discussion
The literature survey reveals significant advancements in sentiment analysis driven by both traditional machine
learning and deep learning techniques. Traditional models, such as Naive Bayes and Support Vector Machines, are
effective in certain contexts, particularly when paired with robust feature selection methods like n-grams and TF-IDF
representations. These methods are foundational in preprocessing, which includes tokenization, stop-word removal,
and stemming. However, deep learning approaches, including Convolutional Neural Networks (CNN) and Long Short-
Term Memory (LSTM) networks, have demonstrated superior performance due to their ability to handle complex, high-
dimensional data. The use of pre-trained word embeddings, such as word2vec and GloVe, has significantly enhanced
these models by providing rich semantic information. Advanced techniques like attention mechanisms and hierarchical
models have further improved sentiment analysis by enabling a more nuanced understanding of textual data. The
survey indicates that model selection and preprocessing techniques must be tailored to the specific dataset
characteristics, including language and domain-specific nuances. Overall, while traditional machine learning models are
valuable for their simplicity and effectiveness, deep learning models are leading the way in achieving higher accuracy
and robustness, showing great promise for future advancements in sentiment analysis across various domains and
languages.

4. Conclusion
In this paper we present a comparative analysis of state-of-the-art systems for sentiment analysis, The analysis
techniques highlight the strengths and weaknesses of both traditional machine learning and deep learning approaches.
Traditional methods like Naive Bayes and Support Vector Machines have shown reliable performance, especially with
well-curated feature extraction techniques. However, the advent of deep learning models such as CNNs and LSTMs has
revolutionized sentiment analysis, offering superior accuracy and the ability to process complex and voluminous data
more effectively. The incorporation of pre-trained embeddings like word2vec and GloVe has further enhanced model
performance by providing a richer semantic understanding of the text. Despite these advancements, it is evident that
the choice of model and preprocessing steps must be contextually adapted to the specific dataset and language
intricacies to achieve optimal results.

In the future, the refinement of deep learning models will be essential to handle the growing complexity of language and
sentiment expression in sentiment analysis. Integrating sophisticated attention mechanisms and hierarchical models
can better capture nuanced relationships within text data. Expanding research into multilingual sentiment analysis and
code-mixed language processing will address challenges in diverse linguistic contexts. Developing real-time
applications for dynamic environments, such as social media monitoring and customer feedback systems, will require

11
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

models to quickly adapt to new data. Incorporating domain-specific knowledge can enhance the accuracy of sentiment
analysis in specialized fields like healthcare, finance, and legal analysis. Addressing ethical considerations and biases in
sentiment analysis models will be crucial to ensure fair and unbiased decision-making.

References
[1] V. Yeole, P. V. Chavan, and M. C. Nikose, Opinion mining for emotions determination, ICIIECS 2015 - 2015 IEEE
Int. Conf. Innov. Information, Embed. Commun. Syst., 2015.
[2] B. Heredia, T. M. Khoshgoftaar, J. Prusa, and M. Crawford, Cross-Domain Sentiment Analysis: An Empirical
Investigation, 2016 IEEE 17th Int. Conf. Inf. Reuse Integr., pp. 160165, 2016.
[3] F. Luo, C. Li, and Z. Cao, Affective-feature-based sentiment analysis using SVM classifier, 2016 IEEE 20th Int. Conf.
Comput. Support. Coop. Work Des., pp. 276281, 2016.
[4] M. Haenlein and A. M. Kaplan, An empirical analysis of attitudinal and behavioral reactions toward the
abandonment of unprofitable customer relationships, J. Relatsh. Mark., vol. 9, no. 4, pp. 200228, 2010.
[5] Shatte, Adrian BR, Delyse M. Hutchinson, and Samantha J. Teague. Machine learning in mental health: a scoping
review of methods and applications. Psychological medicine 49.9 (2019): 1426-1448.
[6] Amini, Mahyar and Ali Rahmani. How Strategic Agility Affects the Competitive Capabilities of Private Banks.
International Journal of Basic and Applied Sciences 10.01 (2023): 8397-8406.
[7] Carrio, Adrian, et al. A review of deep learning methods and applications for unmanned aerial vehicles. Journal
of Sensors 2017 (2017).
[8] Amini, Mahyar and Ali Rahmani. Achieving Financial Success by Pursuing Environmental and Social Goals: A
Comprehensive Literature Review and Research Agenda for Sustainable Investment. World Information
Technology and Engineering Journal 10.04 (2023): 1286-1293.
[9] Zhou, Teng, Zhen Song, and Kai Sundmacher. Big data creates new opportunities for materials research: a review
on methods and applications of machine learning for materials design. Engineering 5.6 (2019): 1017-1026.
[10] Amini, Mahyar, and Zavareh Bozorgasl. A Game Theory Method to Cyber-Threat Information Sharing in Cloud
Computing Technology . International Journal of Computer Science and Engineering Research 11.4 (2023): 549-
560.
[11] Deng, Li, and Dong Yu. Deep learning: methods and applications. Foundations and trends® in signal processing
7.3–4 (2014): 197-387.
[12] Nazari Enjedani, Somayeh, and Mahyar Amini. The role of traffic impact effect on transportation planning and
sustainable traffic management in metropolitan regions. International Journal of Smart City Planning Research
12, no. 2023 (2023): 688-700.
[13] Chai, Junyi, et al. Deep learning in computer vision: A critical review of emerging techniques and application
scenarios. Machine Learning with Applications 6 (2021): 100134.
[14] Jahanbakhsh Javidi, Negar, and Mahyar Amini. Evaluating the effect of supply chain management practice on
implementation of halal agroindustry and competitive advantage for small and medium enterprises .
International Journal of Computer Science and Information Technology 15.6 (2023): 8997-9008
[15] Pouyanfar, Samira, et al. A survey on deep learning: Algorithms, techniques, and applications. ACM Computing
Surveys (CSUR) 51.5 (2018): 1-36.
[16] Amini, Mahyar, and Negar Jahanbakhsh Javidi. A Multi-Perspective Framework Established on Diffusion of
Innovation (DOI) Theory and Technology, Organization and Environment (TOE) Framework Toward Supply
Chain Management System Based on Cloud Computing Technology for Small and Medium Enterprises .
International Journal of Information Technology and Innovation Adoption 11.8 (2023): 1217- 1234
[17] Zhou, Jie, et al. Graph neural networks: A review of methods and applications. AI open 1 (2020): 57-81.
[18] Amini, Mahyar and Ali Rahmani. Agricultural databases evaluation with machine learning procedure. Australian
Journal of Engineering and Applied Science 8.6 (2023): 39-50
[19] Wang, Jinjiang, et al. Deep learning for smart manufacturing: Methods and applications. Journal of manufacturing
systems 48 (2018): 144-156.

12
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

[20] Amini, Mahyar, and Ali Rahmani. Machine learning process evaluating damage classification of composites.
International Journal of Science and Advanced Technology 9.12 (2023): 240-250
[21] Baduge, Shanaka Kristombu, et al. Artificial intelligence and smart vision for building and construction 4.0:
Machine and deep learning methods and applications. Automation in Construction 141 (2022): 104440.
[22] Amini, Mahyar, Koosha Sharifani, and Ali Rahmani. Machine Learning Model Towards Evaluating Data gathering
methods in Manufacturing and Mechanical Engineering. International Journal of Applied Science and Engineering
Research 15.4 (2023): 349-362.
[23] Abd Elaziz, Mohamed, et al. Advanced metaheuristic optimization techniques in applications of deep neural
networks: a review. Neural Computing and Applications (2021): 1-21.
[24] Sharifani, Koosha and Amini, Mahyar and Akbari, Yaser and Aghajanzadeh Godarzi, Javad. Operating Machine
Learning across Natural Language Processing Techniques for Improvement of Fabricated News Model.
International Journal of Science and Information System Research 12.9 (2022): 20-44.
[25] Bo Pang and Lillian Lee. 2004. A Sentimental Education: Sentiment Analysis Using Subjectivity Summarization
Based on Minimum Cuts (ACL ’04). Association for Computational Linguistics, USA, 271śes.
DOI:http://dx.doi.org/10.3115/1218955.1218990
[26] Hang Cui, Vibhu Mittal, and Mayur Datar. 2006. Comparative experiments on sentiment classiication for online
product reviews. In AAAI’06: proceedings of the 21st national conference on Artiicial intelligence - Volume 2July
2006 Pages 1265ś1270, pp. 1265ś1270. 1ś6.
[27] Kamal Nigam and Matthew Hurst. 2004. Towards a robust metric of opinion. In AAAI spring symposium on
exploring attitude and afectin text, Vol. 598603.
[28] Yun Niu, Xiaodan Zhu, Jianhua Li, and Graeme Hirst. 2005. Analysis of polarity information in medical text. In
AMIA annual symposium proceedings, Vol. 2005. American Medical Informatics Association, 570.
[29] Ziqiong Zhang, Qiang Ye, Zili Zhang, and Yijun Li. 2011. Sentiment classiication of Internet restaurant reviews
written in Cantonese. Expert Systems with Applications 38, 6 (2011), 7674ś7682.
[30] Thoudam Doren Singh, Telem Joyson Singh, Mirinso Shadang, and Surmila Thokchom. 2021. Review Comments
of Manipuri Online Video: Good, Bad or Ugly. In Proceedings of the International Conference on Computing and
Communication Systems: I3CS 2020, NEHU, Shillong, India, Vol. 170. Springer, 45.
[31] Loitongbam Sanayai Meetei, Thoudam Doren Singh, Samir Kumar Borgohain, and Sivaji Bandyopadhyay. 2021.
Low resource language speciic pre-processing and features for sentiment analysis task. Language Resources and
Evaluation (2021), 1ś23
[32] Prabhakar, E.; Santhosh, M.; Krishnan, A.H.; Kumar, T.; Sudhakar, R. Sentiment analysis of US Airline Twitter data
using new AdaBoost approach. Int. J. Eng. Res. Technol. (IJERT) 2019, 7, 1–6.
[33] Hourrane, O.; Idrissi, N. Sentiment Classification on Movie Reviews and Twitter: An Experimental Study of
Supervised Learning Models. In Proceedings of the 2019 IEEE 1st International Conference on Smart Systems
and Data Science (ICSSD), Rabat, Morocco, 3–4 October 2019; pp. 1–6.
[34] AlSalman, H. An improved approach for sentiment analysis of arabic tweets in twitter social media. In
Proceedings of the 2020 IEEE 3rd International Conference on Computer Applications & Information Security
(ICCAIS), Riyadh, Saudi Arabia, 19–21 March 2020; pp. 1–4.
[35] Saad, A.I. Opinion Mining on US Airline Twitter Data Using Machine Learning Techniques. In Proceedings of the
2020 IEEE 16th International Computer Engineering Conference (ICENCO), Cairo, Egypt, 29–30 December 2020;
pp. 59–63.
[36] Alzyout, M.; Bashabsheh, E.A.; Najadat, H.; Alaiad, A. Sentiment Analysis of Arabic Tweets about Violence Against
Women using Machine Learning. In Proceedings of the 2021 IEEE 12th International Conference on Information
and Communication Systems (ICICS), Valencia, Spain, 24–26 May 2021; pp. 171–176.
[37] Jemai, F.; Hayouni, M.; Baccar, S. Sentiment Analysis Using Machine Learning Algorithms. In Proceedings of the
2021 IEEE International Wireless Communications and Mobile Computing (IWCMC), Harbin, China, 28 June–2
July 2021; pp. 775–779.
[38] Lina Maria Rojas-Barahona. 2016. Deep learning for sentiment analysis. Language and Linguistics Compass 10,
12 (2016), 701ś719.

13
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

[39] Peerapon Vateekul and Thanabhat Koomsubha. 2016. A study of sentiment analysis using deep learning
techniques on Thai Twitter data. In 2016 13th international joint conference on computer science and software
engineering (JCSSE). IEEE, 1ś6.
[40] Geofrey E Hinton, Simon Osindero, and Yee-Whye Teh. 2006. A fast learning algorithm for deep belief nets. Neural
computation 18, 7 (2006), 1527ś1554.
[41] 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 Proceedings of
the 2013 conference on empirical methods in natural language processing. 1631ś1642.
[42] Eva Cetinic, Tomislav Lipic, and Sonja Grgic. 2018. Fine-tuning convolutional neural networks for ine art
classiication. Expert Systems with Applications 114 (2018), 107ś118.
[43] Hannah Kim and Young-Seob Jeong. 2019. Sentiment classiication using convolutional neural networks. Applied
Sciences 9, 11 (2019), 2347.
[44] Qiongxia Huang, Xianghan Zheng, Riqing Chen, and Zhenxin Dong. 2017. Deep Sentiment Representation Based
on CNN and LSTM. In International Conference on Green Informatics (ICGI),pp. 30-33. 30ś33.
[45] Maryem Rhanoui, Mounia Mikram, Siham Yousi, and Soukaina Barzali. 2019. A CNN-BiLSTM model for document-
level sentiment analysis. Machine Learning and Knowledge Extraction 1, 3 (2019), 832ś847.
[46] K Shalini, Aravind Ravikurnar, Aravinda Reddy, KP Soman, et al. 2018. Sentiment analysis of indian languages
using convolutional neural networks. In 2018 International Conference on Computer Communication and
Informatics (ICCCI). IEEE, 1ś4.
[47] Subhra Jyoti Baroi, Nivedita Singh, Ringki Das, and Thoudam Doren Singh. 2020. NITS-Hinglish-SentiMix at
SemEval-2020 Task 9: Sentiment Analysis for Code-Mixed Social Media Text Using an Ensemble Model. (Dec.
2020), 1298ś1303. https://aclanthology.org/2020. semeval-1.175
[48] Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, and Eduard Hovy. 2016. Hierarchical attention
networks for document classification. In Proceedings of the 2016 conference of the North American chapter of
the association for computational linguistics: human language technologies. 1480ś1489.
[49] Guozheng Rao, Weihang Huang, Zhiyong Feng, and Qiong Cong. 2018. LSTM with sentence representations for
document-level sentiment classiication. Neurocomputing 308 (2018), 49ś57.
[50] Ramadhani, A.M.; Goo, H.S. Twitter sentiment analysis using deep learning methods. In Proceedings of the 2017
IEEE 7th International Annual Engineering Seminar (InAES), Yogyakarta, Indonesia, 1–2 August 2017; pp. 1–4.
[51] Demirci, G.M.; Keskin, ¸S.R.; Do˘gan, G. Sentiment analysis in Turkish with deep learning. In Proceedings of the
2019 IEEE International Conference on Big Data, Honolulu, HI, USA, 29–31 May 2019; pp. 2215–2221.
[52] Raza, G.M.; Butt, Z.S.; Latif, S.; Wahid, A. Sentiment Analysis on COVID Tweets: An Experimental Analysis on the
Impact of Count Vectorizer and TF-IDF on Sentiment Predictions using Deep Learning Models. In Proceedings of
the 2021 IEEE International Conference on Digital Futures and Transformative Technologies (ICoDT2),
Islamabad, Pakistan, 20–21 May 2021; pp. 1–6.
[53] Dholpuria, T.; Rana, Y.; Agrawal, C. A sentiment analysis approach through deep learning for a movie review. In
Proceedings of the 2018 IEEE 8th International Conference on Communication Systems and Network
Technologies (CSNT), Bhopal, India, 24–26 November 2018; pp. 173–181.
[54] Harjule, P.; Gurjar, A.; Seth, H.; Thakur, P. Text classification on Twitter data. In Proceedings of the 2020 IEEE 3rd
International Conference on Emerging Technologies in Computer Engineering: Machine Learning and Internet of
Things (ICETCE), Jaipur, India, 7–8 February 2020; pp. 160–164.
[55] Uddin, A.H.; Bapery, D.; Arif, A.S.M. Depression Analysis from Social Media Data in Bangla Language using Long
Short Term Memory (LSTM) Recurrent Neural Network Technique. In Proceedings of the 2019 IEEE
International Conference on Computer, Communication, Chemical, Materials and Electronic Engineering
(IC4ME2), Rajshahi, Bangladesh, 11–12 July 2019; pp. 1–4.
[56] Alahmary, R.M.; Al-Dossari, H.Z.; Emam, A.Z. Sentiment analysis of Saudi dialect using deep learning techniques.
In Proceedings of the 2019 IEEE International Conference on Electronics, Information, and Communication
(ICEIC), Auckland, New Zealand, 22–25 January 2019; pp. 1–6.
[57] Yang, Y. Convolutional neural networks with recurrent neural filters. arXiv 2018, arXiv:1808.09315.

14
 International Journal of Science and Research Archive, 2024, 12(02), 005–015

[58] Goularas, D.; Kamis, S. Evaluation of deep learning techniques in sentiment analysis from Twitter data. In
Proceedings of the 2019 IEEE International Conference on Deep Learning and Machine Learning in Emerging
Applications (Deep-ML), Istanbul, Turkey, 26–28 August 2019; pp. 12–17.
[59] Chowdhury, S.; Rahman, M.L.; Ali, S.N.; Alam, M.J. A RNN Based Parallel Deep Learning Framework for Detecting
Sentiment Polarity from Twitter Derived Textual Data. In Proceedings of the 2020 IEEE 11th International
Conference on Electrical and Computer Engineering (ICECE), Dhaka, Bangladesh, 17–19 December 2020; pp. 9–
12.
[60] Vimali, J.; Murugan, S. A Text Based Sentiment Analysis Model using Bi-directional LSTM Networks. In
Proceedings of the 2021 IEEE 6th International Conference on Communication and Electronics Systems (ICCES),
Coimbatore, India, 8–10 July 2021; pp. 1652–1658.
[61] Anbukkarasi, S.; Varadhaganapathy, S. Analyzing Sentiment in Tamil Tweets using Deep Neural Network. In
Proceedings of the 2020 IEEE Fourth International Conference on Computing Methodologies and Communication
(ICCMC), Erode, India, 11–13 March 2020; pp. 449–453.
[62] Kumar, D.A.; Chinnalagu, A. Sentiment and Emotion in Social Media COVID-19 Conversations: SAB-LSTM
Approach. In Proceedings of the 2020 IEEE 9th International Conference System Modeling and Advancement in
Research Trends (SMART), Moradabad, India, 4–5 December 2020; pp. 463–467.
[63] Hossen, M.S.; Jony, A.H.; Tabassum, T.; Islam, M.T.; Rahman, M.M.; Khatun, T. Hotel review analysis for the
prediction of business using deep learning approach. In Proceedings of the 2021 IEEE International Conference
on Artificial Intelligence and Smart Systems (ICAIS), Coimbatore, India, 25–27 March 2021; pp. 1489–1494.

15