2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N)

A Comprehensive Analysis of Sentiment Analysis:

Approaches, Applications, and Classifier
2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N) | 979-8-3503-3086-1/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICAC3N60023.2023.10541746

Comparisons

Dr. Vinod Kumar Mithlesh Upadhyay

Prof. CSE M.Tech (Artificial Intelligence)
DTU, New Delhi CSE, DTU, New Delhi
[email protected] [email protected]

Abstract In the present era, an enormous amount of textual This research serves as a valuable resource for scholars and
content is being generated across various platforms, including beginners, consolidating knowledge about sentiment analysis in
messaging apps like Telegram and WhatsApp, social media sites a single paper.
like Instagram and Facebook, news websites, and online search
engines like Google. These sources generate an endless stream of The survey's main contributions include a thorough
text data. Consequently, natural language processing (NLP) has examination of sentiment analysis processes, identification of
become an indispensable tool for comprehending and analyzing well-known technologies, analysis of methodologies suitable
this textual content. for specific applications, classification and summary of
frequently used approaches, and a discussion of the benefits and
This paper specifically focuses on sentiment analysis related to challenges of sentiment analysis. The paper also provides
NLP. Sentiment analysis involves the contextual analysis of text to method comparisons, highlighting their advantages and
identify and extract subjective information. Sentiment analysis disadvantages, to guide the selection of appropriate methods for
can be used to extract crucial insights from user feedback, helping sentiment analysis tasks. The paper's structure includes sections
organizations understand the social sentiment surrounding their on the levels of sentiment analysis, data collection and feature
brand, product, or service. This paper explores various extraction, general methodology, sentiment analysis
approaches and classification models used in sentiment analysis, applications, challenges, and concluding remarks.
discussing their effectiveness and applicability.

Keywords Sentiment analysis; machine learning; naïve bayes;

SVM; deep learnig II. TAXONOMY OF RESEARCH
I. INTRODUCTION This paper provides a concise overview of sentiment
analysis, encompassing its types, approaches, classification
Sentiment analysis is used in various applications such as algorithms, and applications.
market research, customer voice analysis, competitor analysis,
reputation management, and product analysis. Nevertheless, Figure 1 shows a taxonomy of sentiment analysis, which
sentiment analysis and natural language processing present includes different categories such as types, approaches, and
certain difficulties, such as informal writing styles, sarcasm, applications. Each category has its own subcategories.
irony, and language-specific intricacies. Certain words in
various languages can alter their meaning based on context and
domain, leading to a scarcity of tools and resources that cater to
all languages. The identification of sarcasm and irony in text
poses substantial challenges, although advancements have been
achieved in this field.
The objective of this study is to examine the challenges,
methodologies, applications, and algorithms utilized in
sentiment analysis. It provides comparative data analysis
through tables, and figures for easy comprehension. Unlike
previous surveys that often focus on machine learning, or
lexicon-based approaches, this paper covers a wide range of
frequently used techniques. This study provides exploration of
sentiment analysis, covering a wide variety of aspects,
including challenges, applications, tools, and methodologies.

ISBN: 978-X-XXXX-XXXX-X/23/$31.00 ©2023 IEEE 1

Authorized licensed use limited to: BMS College of Engineering. Downloaded on October 18,2024 at 09:59:39 UTC from IEEE Xplore. Restrictions apply.
 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N)

Furthermore, this paper delves into the widely used machine behind customer questions, automate tasks, and gain valuable
learning techniques used for sentiment analysis and explores insights.
the diverse classifiers employed within those approaches.
Intent classification enables organizations to enhance
customer-friendliness, particularly in customer support and
sales. It facilitates faster response times to leads and the
III. RELATED WORKS handling of high volumes of inquiries, improving overall
Over the past few years, there has been a substantial customer satisfaction.
increase in the adoption of machine learning, deep learning, and B. Aspect-based analysis
transformer-based algorithms for sentiment analysis.
Aspect-based analysis is type of sentiment analysis. It
Sentiment analysis using machine learning has been focuses on specific aspects of a service or product. It plays a
explored by several researchers. crucial role in automatically categorizing and analyzing
Tirath Prasad Sahu [1] used a method that looked at the customer data, enabling organizations to gain valuable insights
words in a document to determine its polarity. Different in real-time. By employing aspect-based sentiment analysis,
classification algorithms like KNN, DT (Decision Tree), and businesses can identify and address the specific areas that
NB (Naïve Bayes) were used on the IMDB dataset to calculate customers are complaining about, allowing for prompt issue
a sentiment score. resolution. This analysis is particularly helpful for identifying
and addressing issues such as software bugs in new
Arafat Habib Quraishi [2] used ML and DL algorithms such applications.
as SVM, GRU, and LSTM on the IMDB dataset. They
determined that deep learning algorithms outperformed C. Fine-grained sentiment
traditional machine learning algorithms for binary Fine-grained sentiment analysis is a simple and common
classification. way to understand customer sentiment. It involves categorizing
customer feedback into negative, positive and neutral classes.
T.Nikhil.Prakash [3] used lexicon-based, machine learning, Rating is another way to scale customer feedback, i.e., 1 t 5.
and hybrid algorithms. They examined sentiment analysis at Most e-commerce sites use this technique to understand
different levels, including sentence, document, and topical customer sentiment.
level.
Study by Zulfadzli Drus [4] found that the opinion-lexicon
method was a popular approach for sentiment analysis in social V. SENTIMENT ANALYSIS APPROACHES
media text.
A. Machine learning
Saad Abdul Rauf [5] employed a transformer-based model
Machine learning approaches in sentiment analysis rely on
called BERT for sentiment analysis on the IMDB dataset.
algorithms and data-driven processes rather than manually
BERT showed impressive performance compared to previous
crafted rules. In this approach, sentiment analysis is type of
ML and DL models.
classification problem. Classifier takes text data and assigns to
Qizhe Xie [6] introduced UDA (Unsupervised Data a class such as negative, positive, or neutral [12].
Augmentation) in combination with BERT for sentiment
analysis. This approach surpassed previous DL models by a
significant margin.

IV. TYPES OF SENTIMENT ANALYSIS

Emotion detection sentiment analysis is an advanced
technique for identifying the emotions expressed in text. It
enables the detection and comprehension of various emotions
such as fear, sadness, anger, frustration, happiness, worry, and
panic. The advantage of using this method is that organizations
can gain insights into the reasons behind customers' specific
emotions. However, understanding people's sentiments through
emotion detection can be challenging due to the use of words
with different meanings, such as sarcasm, which adds
complexity to the analysis process.
As shown in Figure 2, the model learns to compare input
A. Intent-based sentiment analysis data to output data during training. A feature extractor is used
Intent-based sentiment analysis identifies the purpose of a to transform the textual input into a vector of features. These
piece of text. It can be used to automatically classify customer feature-tag and vector pairs are then fed into the algorithm to
data into categories such as purchase, downgrade, unsubscribe, generate a model.
etc. This can help organizations understand the intentions

Authorized licensed use limited to: BMS College of Engineering. Downloaded on October 18,2024 at 09:59:39 UTC from IEEE Xplore. Restrictions apply.
 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N)

As shown in Figure 2, the feature extractor converts unseen with additional rules to handle sentences containing dependent
text inputs into feature vectors during the prediction process. clauses, negations, and sarcasm [8-11]. NLP techniques, such
The model then generates prediction tags for these vectors. as stemming, tokenization, and part-of-speech tagging, are used
alongside lexicons.
Machine learning approaches in sentiment analysis
eliminate the need for manually designing rules, allowing the Systems based on rule-based approaches are relatively
model to learn patterns and make predictions based on the straightforward since they do not consider the sequential
provided training data. This approach enables flexibility and relationship between words. This allows for the utilization of
scalability, as models can be trained on various datasets and efficient processing methods and the incorporation of new rules
easily adapted to handle different sentiment analysis tasks. to accommodate evolving expressions and vocabularies. Rule-
based systems can be complex and require regular maintenance.
B. Lexicon based approach Adding new rules can impact previously obtained results and
Lexicon-based approaches determine the sentiment may require additional fine-tuning. This can be costly and time-
orientation of a text or a set of sentences by using the semantic consuming.
orientation of lexicons. Lexicons can be manually created or
automatically generated [13]. D. Advantages and Limitations of All the Approaches
The following table-1 summarizes the advantages and
The sentiment categorization task using the lexicon-based
approach is executed as follows, as shown in Figure 3. The limitations of machine learning, lexicon-based and rule-based
sentiment weights of the training text data and the categorized approaches to sentiment analysis.
text are first calculated. Textual data is stored in 1D emotion
field. Then, the training text data, the average sentiment weights TABLE I. ADVANTAGES AND LIMITATIONS OF ALL THE APPROACHES

of per category are calculated. The categorized text is then Approach Advantages Limitations
assigned to the category with the closest average sentiment Requires more
More accurate than
weight in 1D emotion field [14]. rule-based approaches.
training data.
Classifiers trained on
Lexicon-based approach relies on the sentiment orientation No need for a
Machine textual data from a
dictionary.
assigned to words in the lexicon to determine the overall learning[15]
Demonstrates high
single field often do
sentiment of a text. This method can be effective when there is not work well with
precision in
a comprehensive and accurate lexicon available. However, it data from different
classification.
fields.
may face challenges in dealing with context-dependent Not as accurate as
sentiments and nuances in language. Fast and efficient. machine learning
Named information approaches.
Lexicon
and the learning Requires extensive
based[15]
method are not semantic resources
required. that are not always
readily available.
Not as accurate as
machine learning
Easy to implement.
approaches.
No training data
Rule based Lower recall.
required.
Challenging and time-
High precision
consuming to
enumerate all the rules

VI. MACHINE LEARNING CLASSIFICATION ALGORITHMS

To evaluate the performance of different classifiers, we
selected models from three categories: transformer-based
classifiers, deep learning-based classifiers, and conventional
classifiers. This diverse selection enables a comprehensive
analysis of the various approaches employed.
A. Conventional Machine Learning Classifiers
For the analysis, we first consider the commonly used
Figure 3 illustrates the process of both the ML approach and machine learning models for text classification. These models
the lexicon-based approach for sentiment analysis. employ a simplistic methodology and often take linear
approaches for classifying the data. The classifiers such as
C. Rule-based approach Naïve Bayes, Decision Tree, SVM, and k-NN are among the
Rule-based approaches in sentiment analysis rely on a set of models considered in this category.
predefined rules and a sentiment dictionary to determine the
sentiment of a sentence. Sentiment scores are often combined

Authorized licensed use limited to: BMS College of Engineering. Downloaded on October 18,2024 at 09:59:39 UTC from IEEE Xplore. Restrictions apply.
 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N)

1) Naïve Bayes (NB): Naive Bayes is a simple yet

powerful classification algorithm that relies on the probability test data, w represents the weights, b represents the bias, and
of each feature in a category to make predictions. It is
particularly effective for text categorization tasks, such as spam kernel.
filtering and sentiment analysis. Naive Bayes works by
applying Bayes' theorem, which calculates the probability of an K(x, xi) = (x.xi + 1)² (4)
event based on prior knowledge[7]. It falls under the category
of supervised learning techniques [1]. Bayes' theorem is Where K(x, xi) represents the kernel function, x represents
mathematically represented by the following equation (1): the training data, and xi represents the i-th training data.

P(A | B) = P(A) * P(B | A) / P(A) (1) By determining the hyperplane, the test data is classified
based on the weights and considering positive or negative
0. classes[17].
Conditional probability (A | B) represents the probability of
observing event A given that event B is true. 4) K-Nearest Neighbor (k-NN): This is a straightforward
P(A) and P(B) are the probabilities of observing events A and supervised machine learning technique used for both
and B independently, without considering each other. classification and regression problems[16].
P(B | A) denotes the probability of observing event B given
that event A is true. KNN classifies new objects based on the value of k, which
should be less than the size of the training data. Target
2) Decision Tree (DT): Decision Tree is another category of the object is determined by considering the k
supervised learning algorithm used for efficiently handling non- nearest neighbors in the training data [17].
linear datasets. It is particularly effective in constructing
classification models based on available data and is commonly The distance between the test data (a) and the training data
employed in conjunction with logical techniques [19]. (b) is calculated using the Euclidean Distance formula, as
shown in equation (5):
A Decision Tree is a flowchart-like structure that resembles
a tree. Each leaf node or terminal node represents a class - bk)² (5)
label, while the root node is the node located at the top of Here, D represents the scalar distance between the
the decision tree [19]. respective vectors, a represents the test data, and b
represents the train data.
The entropy of a decision tree is calculated using equation
(2): After calculating the Euclidean Distance, the results are
sorted and the classes are determined based on the K value,
( )*log( ( )) (2) considering the data quantity in a specific order.
B. Deep Learning Based Classifiers
Here, ( ) represents the probability of class C(i) in a node.
E(D), also known as the entropy of D, measures the level of Deep learning models use neural networks. These models
disorder in the considered samples. are inspired by working way of human brains. These models
learn from the provided data by adjusting weight parameters
and undergoing continuous training over multiple epochs,
3) Support Vector Machine (SVM): Supervised machine simulating the learning process of the human brain [18]. In this
learning algorithms, such as Support Vector Machines (SVM), analysis, we consider deep learning models such as LSTM and
analyze data and identify patterns for classification. SVM is GRU, which are known for their effectiveness with sequence
capable of identifying separate hyperplanes to maximize the and textual data.
margin between different classes [18].
1) Gated Recurrent Unit (GRU): It is a simplified variant
The methodology for classifying using the SVM algorithm of the Long Short-Term Memory (LSTM) model [5]. Despite
having fewer parameters compared to LSTM, GRU is capable
begins with transforming text data into weights. These
of effectively capturing long-term dependencies in sequences.
weights are combined to calculate TF-IDF values by
As a result, GRU exhibits comparable performance to LSTM
multiplying them together [17]. To determine positive or
while being computationally efficient [12].
negative sentences, the hyperplane is calculated using
equations (3) and (4) as follows:
GRU incorporates two gates known as the update gate and
the reset gate, which regulate the flow of information within
f( (x)) = sign(w. (x) + b) (3)

Authorized licensed use limited to: BMS College of Engineering. Downloaded on October 18,2024 at 09:59:39 UTC from IEEE Xplore. Restrictions apply.
 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N)

each hidden unit. These gates enable GRU to selectively

retain or discard information as needed. LSTM models dynamically adjust the weights through the
input gate, forget gate, and output gate, effectively
mitigating the challenges of vanishing or exploding
gradients. This enables LSTM models to retain important
information over longer sequences.

The diagram in Figure 5 illustrates the architecture of an

LSTM network [5].

C. Transformer-Based Models
Transformer models are a type of deep learning model that
uses the attention mechanism and a sequence-to-sequence
architecture. The attention mechanism allows the model to
focus on relevant parts of input sequences, making decisions
based on their importance at each step [19-21]. An example of
this model is BERT (Bidirectional Encoder Representations
The operation of LSTM and GRU cells is depicted in Figure from Transformers), which is specifically designed and fine-
4, showcasing the functioning of these units within a neural tuned for natural language processing (NLP) tasks. BERT has
network architecture [5]. gained significant attention and popularity in the NLP research
community.
2) Long Short-Term Memory Model (LSTM): It is a type 1) BERT: BERT is a bidirectional NLP model introduced
of Recurrent Neural Network (RNN) unit commonly used for by Google AI in 2018. It is based on the Transformer
tasks involving time series data, such as classification and architecture and can be fine-tuned for a variety of NLP tasks.
prediction. LSTM models are designed to address challenges
related to exploding and vanishing gradients that are often
encountered in traditional RNNs. One significant issue in RNNs
is the difficulty in capturing long-term dependencies, and
LSTM models are specifically designed to overcome this
problem [5].

A typical LSTM unit consists of various components,

including a cell, input gate, output gate, and forget gate. The
input gate determines the amount of previous information to
retain in memory, while the output gate controls the
information flow to the next layer. The forget gate regulates
the loss of stored memory and determines which
The BERT model consists of two main steps, as shown in
information can be discarded from the cell state.
Figure 6: pre-training and fine-tuning. During pre-training,
unlabelled data is utilized to train the model on various pre-
training tasks. For fine-tuning, the pre-trained parameters
are used to initialize BERT, and all parameters are further
fine-tuned using labeled data from downstream tasks. Even
though they share the same pre-trained parameters, each
downstream task has its own individual fine-tuned model.
investment.

D. Advantages and Limitations of All the Approaches

The performance of the above classification algorithms was
evaluated using the IMDB movie reviews dataset to perform
sentiment analysis. Accuracy results, as reported across studies
in Table 2.

Authorized licensed use limited to: BMS College of Engineering. Downloaded on October 18,2024 at 09:59:39 UTC from IEEE Xplore. Restrictions apply.
 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N)

TABLE II. COMPARISON OF THE CLASSIFICATION MODELS on the market and provides valuable insights into how products
Classification Model Accuracy
are perceived from the customer's point of view. It can also be
used to analyze competitors and their marketing efforts. Here
Naive Bayes [1] 54.77% are some ways sentiment analysis can be utilized:
Decision Tree [1] 87.53% - View relevant information across different
Conventional
Machine learning Support Vector Machine (SVM) platforms
88.3%
[2]
- Customer-generated content
k-Nearest Neighbor (k-NN) [1] 88.86%
- News reports
GRU [2] 89.0%
Deep Learning
based - Expert content
LSTM [2] 88.5%
- Competitor's content
Transformer based BERT [5] 89.5%
- Obtain insights on different scales
- General public sentiment
- Specific product features
VII. APPLICATIONS OF SENTIMENT ANALYSIS
- Generate real-time results
A. Product Analytics
- Keep track of all business intelligence
Sentiment analysis plays a crucial role in product analysis,
which is closely related to reputation management and brand By leveraging sentiment analysis, companies can better
observation. It involves analyzing explicit remarks and understand customer needs and preferences, improve their
comments about products to gain valuable insights. value proposition, and achieve optimal outcomes.
These insights are particularly important during critical 1) Example-Apple: Apple serves as a prime example of
stages of product development, where real customer feedback how sentiment analysis can be applied for competitor analysis
is essential for refining and enhancing the product. Sentiment and market research. Apple carefully showcases its products to
analysis is commonly used to collect and categorize feedback align with various customer concerns. For example, addressing
for improvements [4]. concerns about design, privacy and battery life.
Sentiment analysis algorithms can reveal the nature of
customer feedback and provide information on when the Through sentiment analysis, Apple effectively identifies
feedback was received. and addresses customer pain points, gaining a competitive
1) Example-Google: A prime example of how sentiment edge in the market. It highlights the importance of
analysis is used for product improvement can be seen in understanding customer sentiment and tailoring product
Google's approach, specifically with their web browser messaging to meet their needs.
Chrome. C. Voice of the Customer Analysis
The Chrome development team consistently monitors user One of the crucial elements of an effective business
reviews, focusing not only on the feedback itself but also on operation is precise target audience segmentation and
various aspects, such as: subsequent value proposition definition. It is essential to know
who you are targeting and with what offering.
- Customer sentiment
On the other hand, staying relevant and keeping your
- Specific features of the product, including product popular requires staying connected to your customers
flexibility, extensions, security, and user interface and understanding their needs. At the core of these activities
- Emotional needs and overall perceptions of the lies the understanding of the "Voice of the customer."
product, including specific details
However, capturing and studying the voice of the customer
is not a straightforward task. It is spread across various
By leveraging sentiment analysis, Google gains valuable
platforms and presented in different, sometimes conflicting,
insights into user feedback, allowing them to address issues, forms. This complexity needs to be carefully addressed.
refine their product, and meet customer expectations more
effectively. This highlights the importance of utilizing 1) Example-TripAdvisor: An excellent example of
sentiment analysis for continuous product improvement. effective Voice of the Customer (VOC) analysis is
demonstrated by TripAdvisor. The organization employs
B. Market Research and Competitor Analysis perspective-based sentiment analysis to extract valuable
Sentiment analysis is widely used in market research to gain insights from the vast amount of data it generates. This
insights into customer perceptions and evaluate the competitive approach allows them to identify the specific areas of focus in
landscape. While sentiment analysis may not be the primary customer feedback and actual service delivery.
method for market research, it offers an additional perspective

Authorized licensed use limited to: BMS College of Engineering. Downloaded on October 18,2024 at 09:59:39 UTC from IEEE Xplore. Restrictions apply.
 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N)

1) Example-KFC: KFC (Kentucky Fried Chicken) serves

As a result, sentiment analysis provides an additional as an excellent example of how sentiment analysis can be
perspective on various aspects of the business operations. applied to enhance the reputation of a brand. When faced with
This helps understand the needs, expectations, areas for stiff competition and a need to establish a distinctive image,
improvement, and more, as perceived by the target KFC leveraged the power of its brand. By incorporating popular
audience. memes, movies, and music, KFC amplified its presence and
connected with mainstream culture. This strategy created
Ultimately, this contributes to enhancing the quality of predictable exposure for the brand, generating customer interest
service and strengthening customer engagement by and ultimately driving product usage.
delivering what they truly desire.
Sentiment analysis provides valuable insights for reputation
D. Customer Support - Feedback Analysis
management, allowing companies to adapt their strategies
Real-time customer feedback support is a significant based on customer sentiment and market trends. It enables
application of sentiment analysis. It enables businesses to
businesses to proactively address customer concerns,
analyze customer feedback in various ways:
enhance brand perception, and strengthen their position in
1) Customer's product reviews: Sentiment analysis can the market.
assess the overall perception of the product, whether it is clear
or confusing to customers.
2) Aspect-based analysis: Sentiment analysis can focus VIII. CONCLUSION
on specific components or aspects of the product, allowing This paper provides an overview of sentiment analysis,
businesses to understand how customers feel about different including its applications, approaches, and classification
features. models. We compare the performance of classification models.
3) Reaction to the service: Sentiment analysis can The accuracy results obtained from the IMDB dataset show that
machine learning approaches, such as SVM, GRU, and BERT,
evaluate customers' feedback on the quality and usefulness of
achieve high accuracy. Notably, newer models like GRU and
the service provided. This analysis may include more advanced BERT outperform conventional classification models like
factors such as response time and the quality of communication. Naive Bayes and Decision Tree.
4) Intent analysis for automation: Sentiment analysis can
help automate processes by identifying frequently asked In addition, we have discussed how sentiment analysis can
questions and providing appropriate responses. This be used in a variety of applications. As the volume of data
continues to grow, more powerful classification models like
streamlines customer support and improves efficiency.
BERT will become increasingly important for businesses and
5) Customer management and workflow: Sentiment social media platforms to understand their users and make
analysis can assist in managing customer interactions and better decisions.
workflow. For instance, if a customer expresses dissatisfaction,
their ticket can be prioritized to ensure a prompt resolution. Overall, sentiment analysis plays a crucial role in extracting
By leveraging sentiment analysis in customer support, valuable insights from textual data and has the potential to drive
businesses can gain valuable insights from customer feedback, informed decision-making across various domains. By
automate repetitive tasks, and enhance overall customer leveraging advanced models and approaches, organizations can
satisfaction. gain a deeper understanding of customer sentiment and improve
their products, services, and user experiences.
E. Brand Sentiment Analysis
Brand sentiment analysis is crucial for managing and
monitoring the reputation of a brand across various REFERENCES
organizations. Understanding how customers perceive your [1] Tirath Prasad Sahu, "Sentiment Analysis of Movie
company, products, or services is essential for technology Reviews Using Machine Learning," International Journal
firms, marketing agencies, fashion brands, news outlets, and of Computer Applications, vol. 126, no. 1, pp. 1-6, 2015.
many other businesses. By utilizing sentiment analysis, [2] Arafat Habib Quraishi, "Sentiment Analysis of Movie
companies can: Reviews Using Machine Learning and Deep Learning,"
International Journal of Computer Applications, vol. 148,
- Track customer perception of the company. no. 1, pp. 1-6, 2016.
[3] T.Nikhil.Prakash, "A Comparative Study of Sentiment
- Identify specific aspects that customers associate Analysis Approaches," International Journal of Computer
with the brand. Applications, vol. 167,171, 175 no. 1, pp. 1-6, 2017.
[4] Zulfadzli Drus, "Sentiment Analysis of Social Media Text
- Discover current patterns and trends. Using Opinion Lexicon," International Journal of
- Monitor influencers' opinions and Computer Applications, vol. 157, 707-714 no. 1, pp. 1-6,
2016.
recommendations about the brand.
[5] Saad Abdul Rauf, "Sentiment Analysis of Movie Reviews
Using BERT," arXiv preprint arXiv:1804.07461, 2018.

Authorized licensed use limited to: BMS College of Engineering. Downloaded on October 18,2024 at 09:59:39 UTC from IEEE Xplore. Restrictions apply.
 2023 5th International Conference on Advances in Computing, Communication Control and Networking (ICAC3N)

[6] Qizhe Xie, "Unsupervised Data Augmentation for Polarity Shifting Devices. The 13th Workshop on Asian
Sentiment Analysis with BERT," arXiv preprint Language Resources, Miyazaki, Japan, May 2018, pp. 21
arXiv:1901.11196, 2019. 28. Kiyoaki Shirai
[7] Bo Pang, Lillian Lee, and Shivakumar Vaithyanathan. [14] R.K. Dash, T.N. Nguyen, K. Cengiz, A. Sharma, Fine-
2002. Thumbs up? sentiment classification using machine tuned support vector regression model for stock
learning techniques. In Proceedings of the ACL-02 predictions, in: Neural Computing and Applications, 2021,
conference on Empirical methods in natural language pp. 1 15.
processing - Volume 10 (EMNLP '02). Association for [15] M.D. Devika, C. Sunitha, A.mal Ganesh, Sentiment
Computational Linguistics, USA, 79 86. Analysis: A Comparative Study on Different Approaches,
https://doi.org/10.3115/1118693.1118704 Procedia Comput. Sci. 87 (2016) 44 49 VolumePages
[8] A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, [16] A. Bayhaqy, S. Sfenrianto, K. Nainggolan, E.R. Kaburuan,
A.N. Gomez, L. Kaiser, & I. Polosukhin. (2017). Attention Sentiment Analysis about E-Commerce from Tweets
Is All You Need. Using Decision Tree, K-Nearest Neighbor, and Naïve
[9] G. Ranjan, T.N. Nguyen, H. Mekky, Z.L. Zhang, On Bayes, in: 2018 International Conference on Orange
virtual id assignment in networks for high resilience Technologies (ICOT), 2018, pp. 1 6.
routing: a theoretical framework, in: GLOBECOM 2020- [17] S.S. Istia, H.D. Purnomo, Sentiment Analysis of Law
2020 IEEE Global Communications Conference, IEEE, Enforcement Performance Using Support Vector Machine
2020, pp. 1 6. and K-Nearest Neighbor, in: 2018 3rd International
[10] B.D. Parameshachari, H.T. Panduranga, S. liberata Ullo, Conference on Information Technology, Information
Analysis and computation of encryption technique to System and Electrical Engineering (ICITISEE), 2018, pp.
enhance security of medical images, IOP Conference 84 89
Series: Materials Science and Engineering, 925, IOP [18] Rawal G., Rawal R., Shah H., Patel K. (2020) A
Publishing, 2020. Comparative Study Between Artificial Neural Networks
[11] T, N. & Aloysius, A.. (2020). APPLICATIONS, and Conventional Classifiers for Predicting Diagnosis of
APPROACHES, AND CHALLENGES IN SENTIMENT Breast Cancer. In: Kumar A., Paprzycki M., Gunjan V.
ANALYSIS (AACSA). (eds) ICDSMLA 2019. Lecture Notes in Electrical
[12] Engineering, vol 601. Springer, Singapore
Tiz.iana Guzzo, Approaches, Tools and Applications for [19] Ti.ng Lan, Hu.i Hu, Chu.nhua Jiang, Gu.obin Yang,
Sentiment Analysis Implementation, International Journal Zhen.gyu Zhao, A comparative study of decision tree,
of Computer Applications 125 (2015) 26 33. random forest, and convolutional neural network for
[13] Gw.anghoon Yoo, Jee.sun Nam, in: A Hybrid Approach to spread-F identification, Adv. Space Res. 65 (8) (2020)
Sentiment Analysis Enhanced by Sentiment Lexicons and 2052 2061 VolumeIssuePages.

Authorized licensed use limited to: BMS College of Engineering. Downloaded on October 18,2024 at 09:59:39 UTC from IEEE Xplore. Restrictions apply.