Sentiment Analysis Using Various Tools and

Techniques: A Review
Nikhil Kumarline Jatin Suteri Amit Vishkarma
Department of CSE Department of CSE Department of CSE
Galgotias University Galgotias University Galgotias University
Greater Noida, UP, India Greater Noida, UP, India Greater Noida, UP, India
nikhil.21SCSE1120001@galgotiasuniv [email protected] amit.21SCSE1420044@galgotiasuniver
ersity.edu.in sity.edu.in

Pulkit Kumar Subash Harizan Akhelindra Kgare

Department of CSE Department of CSE Department of CSE
Galgotias University Galgotias University Galgotias University
Greater Noida, UP, India Greater Noida, UP, India Greater Noida, UP, India
[email protected] [email protected] [email protected]

Abstract—With the advancement of technology and World trained and used by businesses for their benefits. This
Wide Web, people have been able to express their thoughts to study's main objective is to provide a comprehensive
everyone. The sentiments of individuals have now gained a new overview of different research papers by different authors
medium to spread, which has resulted in people expressing and viewing from their point of perspective what or how a
their opinions/thoughts on something. It is a medium with a sentiment analysis model should be. We will delve into the
huge amount of information where users can not only share research these researchers did, the methodology they used,
their opinions, but read others' opinions, which can affect their the outcome they got and the conclusion they derived. By
opinion as well. This opinion rich information stored in a
medium called “Internet” is an excellent source for companies
analyzing the existing literature and empirical studies, we
and researchers to understand the sentiment of masses, which will assess the performance of different machine learning
lead to the development of sentiment analysis. Therefore, many algorithms, such as support vector machines (SVM), Naïve
researchers have been doing their studies in this field to Bayes, and random forests, in various sentiment analysis
analyze the emotions of the public. This research is useful for tasks.
the public as their sentiment gets noticed and is useful for Furthermore, this paper will shed light on the importance of
businesses as they can do sentiment analysis and by analyzing high-quality labeled datasets and feature engineering used
the sentiment, take strategies accordingly which may benefit by these authors in sentiment analysis. We will explore
them. Our goal in this paper is to study research conducted by different feature representation methods, including bag-of-
the researchers to identify underlying assumptions, highlight
methodological limitations, and propose avenues for future
words, n-grams, word embeddings, and deep learning-based
techniques, highlighting their impact on sentiment
research.
classification accuracy. Moreover, we will discuss the
Keywords—Machine learning, Sentiment Analysis, Algorithms,
Tools and Techniques
challenges associated with training supervised models, such
as data imbalance, domain adaptation, and the interpretation
I. INTRODUCTION (HEADING 1) of model decisions. By investigating the current views of
In today's world, where social media platforms, online engineers and how they propose their model on sentiment
reviews, and customer feedback play a key role in shaping analysis, this research aims to contribute to the ongoing
public opinion, understanding sentiment has become discourse on sentiment analysis using supervised machine
necessary. Sentiment analysis, a subfield of natural language learning. The insights gained from this study can aid
processing (NLP), has become a powerful tool used to researchers, practitioners, and industry professionals in
extract and comprehend human emotions, attitudes, and selecting appropriate algorithms and methodologies for
opinions from textual data. With the growth of information sentiment analysis applications across diverse domains,
on the internet, sentiment analysis has become important to ranging from customer feedback analysis and brand
understand today's world and the opinion of the people. management to social media monitoring and public
Supervised machine learning approaches have gained sentiment tracking.
significant attention and proven to be highly effective in The rest of the paper is organized into various sections
addressing sentiment analysis challenges. These methods which discusses about the work done by various researchers
learn from labeled examples, enabling them to generalize in different domains, preprocessing of data, which is
and make predictions on unseen text with impressive simplifying the data into simpler parts which can be more
accuracy. This paper aims to explore the potential of easily comprehended by the algorithms by breaking the
supervised machine learning algorithms in sentiment sentences into keywords, approach we followed for
analysis, uncovering their strengths, limitations, and the sentiment analysis, and the rest of the sections are about
underlying methodologies that make them invaluable in this implementations details and results followed by future work
domain. Quite a lot of data is available on the internet which and paper's implementation in daily life. We conducted a
is a particularly reliable source of dataset for these learning comprehensive analysis of 40 research papers authored by
algorithms. Some of the sites that provides important different individuals to gain a comprehensive understanding
datasets are Twitter, YouTube, IMDB movie review, of their studies.
Amazon etc. Excluding these websites, certain sites provide
datasets for these algorithms so that these models can be

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

 II. LITERATURE SURVEY In paper [4] authors, presented a novel approach that uses
In [1] authors mainly focuses on developing a system that sentimental aspects focused on the characteristics of the
can determine whether a tweet is “spam” or “ham” and item. The paper used 28,000 user reviews with over 60
evaluate the emotion of the tweet. The data extracted after product types in this dataset. The system performs pre-
preprocessing the tweets are classified using various processing operations like stone-coating, tokenization,
classifiers, namely, decision tree, logistic regression, boxing, deletion of stop-words from the datasets to extract
multinomial naive Bayes, support vector machine, random meaningful information like positivity or negativity. Natural
forest, and Bernoulli naive Bayes for spam detection. it also Language Processing (NPL) removed stop word from
uses Deep learning methods, namely, simple recurrent dataset. Google co-lab was used as an IDE for the
neural network (RNN) model, long short-term memory experimentation. They used the algorithm to create graphs
(LSTM) model, bidirectional long short-term memory of user reviews. If the bar height in the graph towards 1
(BiLSTM) model, and 1D convolutional neural network means the product reviews are positive, and if the bar height
(CNN) model are used for sentiment analysis. The dataset looks down, that means the semantics of that product are
used for the spam detection has a size of 5572, in which negative. Observation in graphs were made using social
4825 ham and 747 spam contents are present. The dataset network Analysis (SNA), which can cover extensive
used for the sentiment analysis has 31015 tweets, in which networks with millions of nodes.
12548 are labeled neutral, 9685 are labeled positive, and In [5] authors proposed the usage of common machine
8782 are labeled negative class. The classification results learning tools and available social media data to predict the
revealed that the features retrieved from tweets can be used price movement of the Bitcoin, Ethereum, Ripple and
to accurately determine the Sentiment Value of tweets. The Litecoin cryptocurrency market movements. The algorithms
SVM classifier achieved a classification accuracy of 70.56% used were neural network (MLP), Support vector machine
and the deep learning model, namely, LSTM, achieved a and Random Forest, and the dataset was taken from Twitter
validation accuracy of 73.81% for the Twitter sentiment and cryptocompare.com public API. The three algorithms
analysis. were applied on the dataset and accuracy was checked in
In [2] authors proposed several approaches have been each case. Litecoin was found to be the most predictable,
developed for sentiment analysis of social media data. This followed by Bitcoin and Ripple, while Ethereum had an
process of sentiment analysis is usually complex and time- accuracy score of less than 0.50. The paper concluded that
consuming due to the huge amount of data and the MLP was the most successful model, as it managed to
requirement to achieve an elevated level of accuracy. Thus, predict market movement prices in all cryptocurrencies and
this paper presents a deep learning approach for sentiment outperforming all other algorithms in 3 out of 4 cases. SVM
analysis of Twitter data on COVID-19 reviews. The was successful in predicting the market for Ripple, Bitcoin
algorithm is based on an LSTM-RNN based network and and Litecoin, while it failed to predict Ethereum's. RF was
improved feature weighting using an attention layer. This able to predict the Bitcoin and the Litecoin markets.
algorithm uses an improved feature transformation Paper [6] aims to bring out the fact that tweets containing all
framework through an attention mechanism. Four class handles related to COVID-19 and WHO have been
labels (sadness, joy, fear, and anger) from publicly available unsuccessful in guiding people around this pandemic
Twitter data published in the Kaggle database were used in outbreak. This paper is not recent as it was published using
this study. Compared to current approaches, the proposed data from August 27, 2020. It focuses on tweets from that
deep learning approach significantly improved the time, so it may not be accurate in this timestamp. The
performance metrics, with a 20% increase in accuracy and dataset was divided into 2 parts, part 1 had around twenty-
10-12% precision, but only 12-13% recall. Out of a total of three thousand most retweeted tweets from Jan 1, 2020, to
179,108 tweets related to COVID-19, tweets with positive, March 23, 2020. Whereas part 2 contains 266688 collected
neutral, and negative sentiment were found to account for between December 2019 and May 2020. Various algorithms
45%, 30%, and 25%, respectively. Overall, the proposed and Services are used to find the sentiments of Tweets, such
deep learning approach is found to be effective and practical as Naive Bayes, Logistic Regression, fuzzy logic,
and can be easily implemented for sentiment classification WordCloud, SIER model, SPSS and many more.
of COVID-19 reviews. Observations were made that the first dataset portrayed
Authors [3] focus is to analyze the customer reviews about mostly negative or neutral as most people expressed their
various restaurants across Karachi. The database is taken displeasure of lockdown. The second dataset showed that
from the SWOT'S guide to Karachi’s restaurants, which is a people worldwide had shown positive sentiment towards the
popular Facebook group. This research performs sentiment disease.
analysis and classifies each comment as positive, negative, In [7], results shows that Covid-19 vaccine hesitancy is
and by using text categorization techniques, comments are gradually decreasing over time, suggesting that societies'
automatically classified according to feedback about food positive opinions on getting vaccinated have gradually
taste, ambiance, service, and value for money. The increased. This suggests that there is a positive feeling about
algorithms used in this research are Naive Bayes Classifier, COVID-19 vaccination. Daily graphs on positive sentiments
Logistic Regression, Support Vector Machine (SVM), and continue to increase, while neutral and negative sentiment
Random Forest. The authors concluded that the maximum graphs decrease. Some individuals voluntarily opt to get
accuracy was achieved by Random Forest algorithm that is, vaccine boosters because of positive feelings about the
above 90% in sentiment analysis and category classification. Covid-19 vaccination
In paper [8] authors reviewed deep learning approaches
applied to various sentiment analysis tasks and their trends
of development, to help researchers depict the current the views of users and deep learning models are all about
progress and address the current problems. The paper uses the prediction or mimic of human mind, so the deep learning
the database of IMDB, SST, Amazon, YouTube etc. to models provide more accuracy than shallow models.
research on the deep learning approaches. This paper This paper [13] conducts computational study of behavior of
discusses deep learning approaches and their applications on a person by checking his buying interest and then finding his
different tasks of sentiment analysis. Those tasks involve opinions about a company’s business entity. It uses datasets
document level, sentence level, ABSA, multi-lingual, and from Amazon which contain reviews of Camera, Laptops,
multimodal sentiment analysis. This paper also highlights Mobile phones, tablets, TVs, video surveillance. After
current issues that need to be addressed and provides preprocessing, the authors applied machine learning
suggestions for improvement including using new signals, algorithms to classify reviews that are positive or negative.
new models like GANs, DRL models, usage of sentiment- The authors use 2 machine learning algorithms: Naive
specific word embedding models and BERT language Bayes and SVM. This paper concludes that Machine
model. Learning Techniques gives the best results to classify the
Paper [9] provides a detailed survey of popular deep Products Reviews. Naïve Bayes got accuracy 98.17% and
learning models used in sentiment learning. The paper Support Vector machine got accuracy 93.54% for Camera
presents a taxonomy of sentiment analysis and application Reviews.
of popular deep learning architectures. The paper discusses This paper [14] introduces an approach to selection of a new
popular deep learning approaches such as CNNs, Rec NNs, feature set based on Information Gain, Bigram, Object-
RNNs, LSTM, GRU, and Deep Belief Networks along with oriented extraction methods in sentiment analysis on social
their architecture, and the important and famous work using networking side. In addition, this paper also proposes a
these architectures in sentiment analysis. The paper explored sentiment analysis model based on Naive Bayes and Support
the primary sentiment analysis dataset, key features of the Vector Machine. Its purpose is to analyze sentiment more
dataset, deep learning model applied to them, and the effectively. The dataset was 500 tweets from Stanford
accuracy (or F1 score) obtained from dataset. The paper Corpus, 200000 tweets from Standfort University, 20000
reviewed around 200 articles and concluded that sentiment tweets of Ravikiran Janardhana. The paper concluded that
analysis using deep learning is a promising research area. authors build a model which analyzes sentiment on twitter
Paper [10] uses the large amount of quality of sentiment rich which applies Bigram, Unigram, Object-oriented features as
data in the form of tweets, status updates, blog posts etc. in an effective feature set for sentiment analysis, and this
Twitter, which is extremely helpful in knowing the opinion model proved to be highly effective and accurate on the
of the crowd. In this paper, authors try to analyze the twitter analysis of feelings.
posts about electronic products like mobiles, laptops etc. This paper [15] conducts experiments by using various
using Machine Learning approach. They used different techniques to identify the polarity of the tweets. It analyzed
classification techniques like naive bayes classifier, SVM the Movie reviews using various techniques like Naïve
classifier, maximum entropy classifier and ensembler Bayes, K-Nearest Neighbour and Random Forest. Data was
classifier to classify the dataset. They presented a new collected from 2000 user-created movie reviews archived on
feature vector for classifying the tweets as positive, negative the IMDb. The dataset was imported in WEKA and then
and extract peoples’ opinion about products. They saved in ARFF format. Filters were applied to the dataset
concluded that all these classifiers have almost similar and 3 algorithms were applied on dataset: Naive Bayes, K-
accuracy for the new feature vector. This feature vector nearest Neighbour and Random Forest. This paper
performs well for electronic products domain. concluded that the best results were given by Naïve Bayes
This paper [11] uses four state-of-the-art machine learning classifier. The Naïve Bayes classifier achieved 81.45%
classifiers (Naïve Bayes, J48, BFTree and OneR) for accuracy, Random Forest classifier achieved 78.65%
optimization of sentiment analysis. This experiment was accuracy, K-Nearest Neighbour classifier achieved 55.30%
performed using three manually compiled datasets, out of accuracy. They also concluded that intelligent systems can
which 2 were picked from amazon and 1 dataset was be developed that can provide users with comprehensive
assembled from IMDB movie reviews. The experiment reviews of movies, products, and services without the need
concluded that The Naïve Bayes was quite fast in learning for the user to go through individual reviews.
whereas OneR seems more promising in generating the This paper [16] presents a system which collects posts from
accuracy of 91.3% in precision, 97% in F-measure and social networking sites, analyzes those Tweets and thus
92.34% in correctly classified instances. They also provides some prediction of business intelligence. It will
concluded that J48 and OneR are better for smaller datasets. help in finding positive, negative, and neutral tweets from
This Review Paper [12] highlights latest studies regarding information resources. The tool proposed in the paper
the implementation of deep learning models such as deep analyzes the current trends in technology and provides a
neural networks, convolutional neural networks, and many way to find out technology trends in the future. The
others for solving different problems of sentiment analysis database used in this paper was multiple tweets on Twitter
such as sentiment classification, cross lingual problems, on the topic "#technology". The dataset was filtered to
textual and visual analysis, and product review analysis. check if there were not any repeating tweets. The algorithms
This paper was a review paper, so it analyzed different deep used were Naive bayes classifier and Maximum Entropy
learning models and the studies conducted on them. The algorithm, which the authors modified and presented as
paper concluded that by using deep learning methods, "Proposed Naive bayes classifier" and "Proposed Maximum
sentiment analysis can be accomplished in a more efficient Entropy classifier". The paper concluded that using
and accurate way. As sentiment analysis is used to predict emoticons as noisy labels for training data is an effective
way to perform distant supervised learning. Machine SA methods to detect fake reviews. SA and text
learning algorithms can achieve high accuracy for classification methods are applied to a dataset of movie
classifying sentiment when using this method. reviews. Five Supervised machine learning algorithms were
This paper [17] uses the data from the tweets in which used: Naive Bayes (NB), Support Vector Machine (SVM),
opinions are highly unstructured. The paper contributes to K-Nearest Neighbors (KNN-IBK), KStar (K*) and Decision
the sentiment analysis for customers’ review classification Tree (DT-J48). Using the accuracy analysis for these five
which is helpful to analyze the information in the form of techniques, they found that SVM algorithm is the most
the number of tweets where opinions are highly unstructured accurate for correctly classifying the reviews in movie
and are either positive or negative, or somewhere in between reviews datasets (approximately 81.35%).
of these two. The dataset used was 19340 Tweets, out of This paper [22] puts forward a study comparing various
which 18340 were used for training and 1000 for testing. Machine learning, Deep learning, and their hybrid
Pre-processing is applied to the dataset which simplifies the techniques. It compares their accuracy for Sentiment
data and makes it simpler for the algorithms to work. 3 Analysis and thus it can be concluded which one is better.
algorithms were used in this experiment: Naive Bayes, The paper explains types of machine learning classifiers
Maximum entropy and SVM. After that Semantics (Naive Bayes, Support Vector Machine (SVM), Decision
orientation-based WordNet was used to extract synonyms Tree, Maximum Entropy) and Deep learning networks
and was used to measure the similarity between content. The (Convolutional Neural Networks, Probabilistic Neural
paper concluded that the naïve byes technique gives us a Network, Recurrent Neural Network, LSTM), which are
better result than the maximum entropy and SVM. Further more complex machine learning implementations. It can be
the accuracy is improved when the semantic analysis concluded that in most cases Deep learning techniques give
WordNet is followed up by the above procedure taking it to better results. However, in some cases the difference in the
89.9% from 88.2%. accuracies of the two techniques is not substantial enough
This paper [18] explores different deep learning algorithms and thus it is better to use Machine Learning methods as
that have not been applied to the Arabic language. The they are easier to implement.
Arabic language imposes many challenges due to its This paper [23] provides a detailed survey on basis of online
complex structure and various dialects. Therefore, the social networks (OSN). They analyse on basis of social
sentiment analysis on Arabic language still needs room for networking sites as in today’s era most people spend most of
improvement. This paper uses an ensemble model their time on online social networks. Thus, they applied
combining Convolutional Neural Network and Long Short- techniques such as monitoring and recommendation systems
Term Memory. These 2 deep learning models were (RS) which help to collect and analyze this data. The
concatenated to form a model which uses the ASTD dataset sentiments and emotions expressed on the messages posted
which consists of 10,000 Arabic tweets. The paper on OSN provide clues and unveil various hidden patterns
concluded that their ensemble model achieved a F1-score of and various aspects of the behavior of users; for instance,
64.46%, which outperforms the state-of-the-art deep sentences containing words. How the user is typing, in what
learning's F1-score of 53.6%. manner, like, dislike and many more which provide lot of
This paper [19] used various classification techniques to data to gain conclusion to understand and based on that
classify the reviews into positive and negative. Over 400000 knowledge. Knowledge-Based Recommendation System
reviews of mobile phones were taken as a dataset. Out of used.
various classification models, Naive Bayes, Support Vector This paper [24] reviews Analyzing sentiment is a process to
Machine (SVM) and Decision Tree were used in find out the opinion of a text. Various to text written on
classification of reviews. The evaluation of models is done various social media platform and a model is drown on that
using 10-Fold Cross Validation. The data was first balanced basis and hidden pattern from text is drown on besis of
with an almost equal ratio of positive and negative People post comments in social media mentioning their
sentiments. They observed that SVM had the best predictive experience about an event as it is always interested to know
accuracy of 81.75%. The authors expect that by working if most other people had a positive or negative experience
more on this model, it can also predict the rating consumers on the same event or what. Analyzing sentiment is a process
gave. of knowing users' emotions for a particular item which may
This paper [20] includes the adoption of a hybrid approach be an event or topic or individual of contemporary trends.
that involves a sentiment analyzer that includes machine Sentiment analysis can be done at three levels, and they are
learning. It also provides a comparison of techniques of sentence, aspect, and document level.
sentiment analysis in the analysis of political views by The purpose [25] of this study was to analyze reaction of
applying supervised machine-learning algorithms such as citizens from diverse cultures to the novel Coronavirus and
Naive Bayes and Support vector machines (SVM). But this people’s sentiment about subsequent actions taken by
research is focused on providing a comparison between different countries. Deep long short-term memory (LSTM)
sentiment lexicons (W-WSD, SentiWordNet, TextBlob) so models used for estimating the sentiment polarity and
that the best can be adopted for sentiment analysis. Authors emotions from extracted social medias such as tweets have
observed that even though the results of TextBlob were been trained to achieve state-of-the-art accuracy on the
better, they obtained the best result when analyzing tweets sentiment140 dataset. How diverse cultures react and
with W-WSD, as is clearly shown in the results. respond to a given crisis is predominant in a society’s norms
This paper [21] aims to classify movie reviews into groups and political will to combat the situation. Often, the
of positive or negative polarity by using machine learning decisions made are necessitated by events, social pressure,
algorithms. Authors analyzed online movie reviews using or the need for the hour, which may not represent the
nation’s will. While some are pleased with it, others might text data on the web. In this regard, opinion data can be used
show resentment. Coronavirus (COVID-19) brought a mix to reveal the sentiments and distribution of the diverse
of similar emotions from the nations towards the decisions groups of people. User opinions expressed on the web affect
taken by their respective governments. Social media was readers in terms of getting an idea on some issues. Similarly,
bombarded with posts containing both positive and negative the opinions expressed by users are especially important for
sentiments on the COVID-19, pandemic, lockdown, and product vendors and politicians. In fact, there are also
hashtags past couple of months. Despite geographically economic effects of this process. Considering all these
close, many neighboring countries reacted differently to one aspects, analysis of the web text data and obtaining
another. For instance, Denmark and Sweden, which share meaningful information are especially important and these
many similarities, stood poles apart on the decision taken by notions have drawn the attention of many researchers.
their respective governments. In this paper [30], various sentiment analysis types and
With the [26] emergence of the development of the Web methods we have studied. Our aim is to produce sentiment
2.0, there is a huge amount of textual content over the Analysis which will efficiently categorize various reviews.
internet including news articles and historical documents, Machine learning methods like SVM and Maximum
with a notable increase after the rise of social media, such as Entropy methods were discussed here in brief, along with
Twitter platform. More people start to express their feelings some other interesting methods that can improve the
and opinions across the internet and various social media. analysis process in one way or the other. Semantic analysis
This led to an increase in the number of user-generated is a great consideration of text analysis. Research work is
sentences containing sentimental information. Investigating carried out for better analysis methods in this area, including
new methods to gain different insight into how people feel semantics by considering n-gram evaluation instead of
and respond to different situations is inevitable. This paper word-by-word analysis. In the world of the internet most
compares the performance of different machine learning and people depend on social networks to get some knowledge
deep learning algorithms, in addition to introducing a new about, they will buy a product. Everyone wants a product
hybrid system that uses text mining and neural networks for first we go to the internet sites to analyze the product and
sentiment classification. after analyzing we buy that product.
This paper [27] conducts experiment about sentimental This paper [31] presents a series of models that use a
analysis on besis of current culture of videos watching combination of automatic manual feature separation
,creating ,playing… in current scenario With the increasing together with a set of analyzes that have been trained
popularity of video sharing websites such as YouTube and accordingly. In addition, a deep learning methodology is
Facebook, multimodal sentiment analysis has received proposed to evaluate the performance of combined work.
increasing attention it provide huge data to understand Six datasets are used in relation to two social sites of
sentiments of people from in what context they have created different domains. Statistical analysis is performed to
video, which type of video is liked at what time frame determine the information combined through various
,about likes ,dislikes ,shares, expression and many more and analyzes and functions sufficient to outperform sentiment
provide sentiment analysis which focus on holistic classification. The proposed work deals with the basic
information in speech segments such as bag of words framework needed to characterize already available
representations and average facial expression intensity, they sentiment analysis based on traditional research
develop a novel deep architecture for multimodal sentiment methodologies with respect to deep learning techniques. The
analysis that performs modality fusion at the word level. analysis also shows that the proposed work shows a
This paper [28] presents a system which collects posts from significant improvement over the already existing
social networking sites, analyzes those Tweets and thus techniques in terms of performance. This shows a set of data
provides some prediction of business intelligence. It will obtained from various sources such as influence word
help in finding positive, negative, and neutral tweets from vectors, generic and surface features, which will lead to
information resources. The tool proposed in the paper positive improvement of sentiment analysis tasks. This
analyzes the current trends in technology and provides a research paper will also conclude about the testing
way to find out technology trends in the future. The dataset methodology that can increase the performance of deep
was filtered to check if there were not any repeating tweets. sentiment analysis. In this paper [32], we compared different
The algorithms used were Naive bayes classifier and sentiment analysis techniques on a dataset collected from
Maximum Entropy algorithm, which the authors modified 900 users from Facebook, along with information on users'
and presented as "Proposed Naive bayes classifier" and age and gender. We extracted this dataset into four groups to
"Proposed Maximum Entropy classifier". The paper analyze the impact of age and gender on the way a user
concluded that using emoticons as noisy labels for training expresses their opinion. Machine learning and dictionary-
data is an effective way to perform distant supervised based techniques were performed to know the sentiment
learning. Machine learning algorithms can achieve high analysis of the reviews. In terms of gender, female data
accuracy for classifying sentiment when using this method. recorded the best accuracy, while for age, the Over 50 group
This paper [29] uses to understand and get hidden patterns has better accuracy compared to all other age groups. The
from reviews, feedback given on various social media results can be further improved by collecting more data for
platforms nowadays, the feedback about the user behavior both men and women and different age groups.
can be achieved by web-based sharing platforms such as
social networks, blogs, forums, and review sites. The III. CONCLUSION
growing popularity of diverse types of personal publishing In this paper, we concluded that supervised machine
services shows that opinions are critical issues in terms of learning has proven to be a valuable approach in sentiment
analysis. We read many papers by different authors and tried [16] Hemalatha, I., Varma, G. S., & Govardhan, A. (2013). Sentiment
to understand the basics behind the working, analysis tool using machine learning algorithms. International
Journal of Emerging Trends & Technology in Computer Science
implementation, and uses of sentiment analysis. We then (IJETTCS), 2(2), 105-109.
tried to summarize their research and observed that the [17] Gautam, G., & Yadav, D. (2014, August). Sentiment analysis of
research was conducted using various labeled training data twitter data using machine learning approaches and semantic analysis.
and employing algorithms like Naive Bayes, Support Vector In 2014 Seventh international conference on contemporary computing
Machines (SVM), and Random Forests. Techniques such as (IC3) (pp. 437-442). IEEE.
bag-of-words and TF-IDF contribute to capturing sentiment- [18] Heikal, M., Torki, M., & El-Makky, N. (2018). Sentiment analysis of
Arabic tweets using deep learning. Procedia Computer Science, 142,
related information from text effectively (pre-processing).
114-122.
Different datasets were used like Tweets from Twitter, [19] Singla, Z., Randhawa, S., & Jain, S. (2017, June). Sentiment analysis
Amazon's product review, IDMB's movie reviews to of customer product reviews using machine learning. In 2017
analyze the sentiment of the comments and train the international conference on intelligent computing and control
supervised model. (I2C2) (pp. 1-5). IEEE.
[20] Hasan, A., Moin, S., Karim, A., & Shamshirband, S. (2018). Machine
REFERENCES learning-based sentiment analysis for twitter accounts. Mathematical
and computational applications, 23(1), 11.
[1] Rodrigues, A. P., Fernandes, R., Shetty, A., Lakshmanna, K., &
Shafi, R. M. (2022). Real-time twitter spam detection and sentiment [21] Elmurngi, E., & Gherbi, A. (2017). Detecting fake reviews through
analysis using machine learning and deep learning sentiment analysis using machine learning techniques. DATA
techniques. Computational Intelligence and Neuroscience, 2022. ANALYTICS, 9.
[2] Singh, C., Imam, T., Wibowo, S., & Grandhi, S. (2022). A deep [22] Jain, K., & Kaushal, S. (2018, August). A comparative study of
learning approach for sentiment analysis of COVID-19 machine learning and deep learning techniques for sentiment analysis.
reviews. Applied Sciences, 12(8), 3709. In 2018 7th International Conference on Reliability, Infocom
[3] Zahoor, K., Bawany, N. Z., & Hamid, S. (2020, November). Technologies and Optimization (Trends and Future
Sentiment analysis and classification of restaurant reviews using Directions)(ICRITO) (pp. 483-487). IEEE.
machine learning. In 2020 21st International Arab Conference on [23] Rosa, R. L., Schwartz, G. M., & Ruggiero, W. V. (2019). Demostenes
Information Technology (ACIT) (pp. 1-6). IEEE. Z. Rodr ıguez, Senior Member, IEEE,“A Knowledge-Based
[4] Wassan, S., Chen, X., Shen, T., Waqar, M., & Jhanjhi, N. Z. (2021). Recommendation System that includes Sentiment Analysis and Deep
Amazon product sentiment analysis using machine learning
Learning”. IEEE Transactions on Industrial Informatics, 15.
techniques. Revista Argentina de Clínica Psicológica, 30(1), 695.
[5] Valencia, F., Gómez-Espinosa, A., & Valdés-Aguirre, B. (2019). [24] Poornima, A., & Priya, K. S. (2020, March). A comparative sentiment
Price movement prediction of cryptocurrencies using sentiment analysis of sentence embedding using machine learning techniques.
analysis and machine learning. Entropy, 21(6), 589. In 2020 6th international conference on advanced computing and
[6] Chakraborty, K., Bhatia, S., Bhattacharyya, S., Platos, J., Bag, R., & communication systems (ICACCS) (pp. 493-496). IEEE.
Hassanien, A. E. (2020). Sentiment Analysis of COVID-19 tweets [25] Imran, A. S., Daudpota, S. M., Kastrati, Z., & Batra, R. (2020).
by Deep Learning Classifiers—A study to show how popularity is Cross-cultural polarity and emotion detection using sentiment
affecting accuracy in social media. Applied Soft Computing, 97, analysis and deep learning on COVID-19 related tweets. Ieee
106754. Access, 8
[7] Qorib, M., Oladunni, T., Denis, M., Ososanya, E., & Cotae, P. [26] Abd El-Jawad, M. H., Hodhod, R., & Omar, Y. M. (2018, December).
(2023). COVID-19 vaccine hesitancy: Text mining, sentiment Sentiment analysis of social media networks using machine learning.
analysis and machine learning on COVID-19 vaccination Twitter In 2018 14th international computer engineering conference
dataset. Expert Systems with Applications, 212, 118715.
(ICENCO) (pp. 174-176). IEEE.
[8] Habimana, O., Li, Y., Li, R., Gu, X., & Yu, G. (2020). Sentiment
[27] Chen, M., Wang, S., Liang, P. P., Baltrušaitis, T., Zadeh, A., &
analysis using deep learning approaches: an overview. Science
China Information Sciences, 63, 1-36. Morency, L. P. (2017, November). Multimodal sentiment analysis
[9] Yadav, A., & Vishwakarma, D. K. (2019). Sentiment analysis using with word-level fusion and reinforcement learning. In Proceedings of
deep learning architectures: a review. Artificial Intelligence Review, the 19th ACM international conference on multimodal
53(6) interaction (pp. 163-171).
[10] Neethu, M. S., & Rajasree, R. (2013, July). Sentiment analysis in [28] Hemalatha, I., Varma, G. S., & Govardhan, A. (2013). Sentiment
twitter using machine learning techniques. In 2013 fourth analysis tool using machine learning algorithms. International
international conference on computing, communications and Journal of Emerging Trends & Technology in Computer Science
networking technologies (ICCCNT) (pp. 1-5). IEEE. (IJETTCS), 2(2), 105-109.
[11] Singh, J., Singh, G., & Singh, R. (2017). Optimization of sentiment [29] Aydoğan, E., & Akcayol, M. A. (2016, August). A comprehensive
analysis using machine learning classifiers. Human-centric survey for sentiment analysis tasks using machine learning
Computing and information Sciences, 7, 1-12. techniques. In 2016 international symposium on innovations in
[12] Ain, Q. T., Ali, M., Riaz, A., Noureen, A., Kamran, M., Hayat, B., intelligent systems and applications (INISTA) (pp. 1-7). IEEE.
& Rehman, A. (2017). Sentiment analysis using deep learning [30] Devika, M. D., Sunitha, C., & Ganesh, A. (2016). Sentiment
techniques: a review. International Journal of Advanced Computer analysis: a comparative study on different approaches. Procedia
Science and Applications, 8(6). Computer Science, 87, 44-49.
[13] Jagdale, R. S., Shirsat, V. S., & Deshmukh, S. N. (2019). Sentiment [31] Pandian, A. P. (2021). Performance evaluation and comparison
analysis on product reviews using machine learning techniques. using deep learning techniques in sentiment analysis. Journal of
In Cognitive Informatics and Soft Computing: Proceeding of CISC Soft Computing Paradigm (JSCP), 3(02), 123-134.
2017 (pp. 639-647). Springer Singapore. [32] Kumar, S., Gahalawat, M., Roy, P. P., Dogra, D. P., & Kim, B. G.
[14] Le, B., & Nguyen, H. (2015). Twitter sentiment analysis using (2020). Exploring impact of age and gender on sentiment analysis
machine learning techniques. In Advanced Computational Methods using machine learning. Electronics, 9(2), 374.
for Knowledge Engineering: Proceedings of 3rd International
Conference on Computer Science, Applied Mathematics and
Applications-ICCSAMA 2015 (pp. 279-289). Springer International
Publishing.
[15] Baid, P., Gupta, A., & Chaplot, N. (2017). Sentiment analysis of
movie reviews using machine learning techniques. International
Journal of Computer Applications, 179(7), 45-49.