Volume 9, Issue 6, June – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24JUN1645

A Review on Sentiment Analysis

Techniques and Approaches
Devendra Singh Rathore1 Dr. Pratima Gautam2
Rabindranath Tagore University Raisen, M.P. India Rabindranath Tagore University Raisen, M.P. India

Abstract:- In current scenario persons share their  In Sentence level Every sentence is treated as an entity
emotions and views on social media in symbolically and in the sentence-level approach, and analyses are
text form. These views turn out to be beneficial to applied to each sentence individually before their
organizations, political institutions and the public. It can results are summed up and the result is supplied to the
be difficult to record and understand consumer user as a result. [4].
emotions because reviews on the Internet are available to  Entity and Aspect level is an approach that is based on
millions for a product or service. Sentiment examination features. During this process, we consider both positive
assumes a significant part in corporate life as they and negative sentiments about features.
influence their dynamic cycle in different sorts of
occasions they face. The essential goal of this survey is to This is an approach that is based on features. During
give a total image of sentiment investigation techniques this process, we consider both positive and negative
and approaches, its sorts and grouping. This research sentiments about features.
paper presents an insight of different approaches on
sentiment analysis along with demerits.  Issues in Sentiment Analysis

Keywords:- Sentiment Analysis, Machine Learning,  Positive and negative sentiment might have inverse
Algorithm, NLP. implications in an exceptionally specific space, so
foreseeing its significance from its keyword is
I. INTRODUCTION troublesome. It is difficult to analyze such peaceful
sentences in terms of positive or negative aspects, but the
Sentiment analysis uses natural language processing in keyword is still of important and importance significance
order to extract useful information about public reactions for dissecting these sentences[2].
to a specific product or topic. Generally, sentiment analysis  Sarcastic Sentences is terms of a style of jocks, very few
is the analysis of feelings and opinions from a sentences may violate the complete meaning of the entire
predetermined point of view. It involves using blog posts, sentences, since quiet sentences need attention to the
comments, and reviews to collect and document the public's keyword and sentence structure. Not only do these funny
attitude toward the product. Increasing interest in e- sentences destroy a sentence, yet, they additionally
commerce is also a significant means of expressing and obliterate the worth of the entire archive. Ex-What an
analyzing opinions, which is an important source of incredible vehicle? It hasn't worked after just 10 days.
sentiment, estimation and reviews [1]. People constantly  Language Issues Change Place to position like Motorola
seek other*s' opinion for decision making on any topic, word is utilized as motto, Lamia word word might be
because opinions are the key to most of their activities utilized as Lummi. Such reasonable phrases alter the
and the crucial influencer of their behavior. There are meaning of the problem and phrases.
different classification of tools for sentiment analysis like  Conditional sentences are a challenge when analyzing
Senti WordNet, Happiness Index, SenticNet, emoticons, feelings. conditional sentences create the same problem
liwc. that interrogative sentences do. Ex-If the image nature
of this camera is delicate, I will be able to go for it.
 Different Level of study  Author & Reader Perspective like The dollar price
There are three types of sentiment analysis: analysis at increases with the Indian currency unit. Concern
the Document level, Sentence level and Aspect or entity document has both meaning the positive and negative
level. along with variation of values from one person to
another person. Sentence having a positive connotation
 Document level is possible that results produced at the to exporter, while the same sentence has a negative
document level are sometimes inappropriate because connotation to the importer.
the whole document is taken into account as one unit.  Spam sentiments are Sentiments posted by an alternative
or competing company to increase their product value or
their company value among customers.

IJISRT24JUN1645 www.ijisrt.com 2148

Volume 9, Issue 6, June – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24JUN1645

II. COMPUTATIONAL APPROACHES

 To carry out Sentiment Analysis, there are two Different Approaches which are as follows:

Fig 1 Approaches of Sentiment Analysis

 Machine Learning III. RELATED WORKS

Machine learning could be a data analytics approach
that automates the analytical model. There are algorithms Mondher Bouazizi Et Al [30] portray a system towards
that are used iteratively to discover from the available data, arranges the messaes that is gathered from different social
which helps the system to appear beyond views with no destinations into numerous opinion classes. In this
outside [10]. In machine learning, the machine learns examination paper analyst arranges tweets into seven
from the preceding data, and on the basis of the preceding classes.
data, it produces reliable decisions making outputs. there
are two styles of machine learning approaches. Wanxiang Che Et Al [31] portray a structure
forsentence pressure strategy include based feeling
 Supervised Learning: In supervised learning, we've got investigation. The specialist portray syntactic pressure
the computer file (x) and output data (y) and an procedure for more effective than extractive pressure
algorithm is understood which must map the functions of method
given data with desired data. In supervised learning there
are again differing kinds of supervised learning like Tang et al. [32] talk about four issues connected with
Decision Tree classification, support vector machine, assessment mining i.e., arrangement of abstract,Order of
Naïve Bayes, Neural network etc feeling by word, grouping of opinion by archive, and
 Unsupervised learning: It is work with no labeled examination of assessment. For the order of subjectivity,
answers. The foremost example of this system is cluster they underlined approaches like ward closeness, the NB
analysis. There are different types of clustering methods classifier, the numerous NB classifier, and the cut-based
like Partitioning method, Density based method, classifier.
Hierarchical method etc.[16]
Lu Lin et al. [33] proposed a structure to take
 Lexicon Based Approach advantage of conclusions in the field of microblogists and
One more way to deal with sentiment examination afterward build a continuous examination framework to
includes the calculation of feeling of the semantic direction screen the spread, variance of sentiments and consuming
of the word or expression that happens in the text. During suppositions in Weibo.
this approach Each word is assigned an emotion score
from both the positive and negative dictionaries. These ElRahman, et al [34] worked on twitter data set to find
words are then compared to the emotional value of the the popularity of res taurant using machine learning
word or sentence. [17] approaches. they mainly focus on polarity to find the
appropriate results.

IJISRT24JUN1645 www.ijisrt.com 2149

Volume 9, Issue 6, June – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24JUN1645

Ramanathan et al [35] focus on opinion or feedback in The information from the paper is extricated and the
tourism mainly in oman tourism. For this they used the essential review discoveries are dissected and coordinated
twitter data set and proposed a lexicon based approach for into a Table I.
finding better reults.

Table 1 Previous Research Papers

Previous Research Papers Summary
S.N. Author Name Tools/Method Used Description Dataset
1 SVM, Na-ive Bayes and The unified model combining the syntaxic context Twitter
Tang et al., 2014) NB-enhanced SVM, of words
Recursive Auto encoder and sentence feeling information gives the best
result in both experiments.
2 Pablo et al (2014) The system uses a basic rule that looks for polarity Twtter
Naïve Bayes words in
the tweets/texts analyzed. It reaches 63 percent of F-
score.
4 The creators propose an open Utilization of ten-fold cross-validation to keep away Twitter
system to gather bias due to
ÁlvaroCuesta et al and investigate information train-ing set partition. Parameters such as variance of
(2014) from Twitter's public accuracy, variance of F-measure,Recall were used
streams consequently. for the assessment.
Naïve Bayes and ngrams
5 The SVM evaluation was performed using Twitter
parameters such as
Bian et al (2015) NLP and SVM the below-curve area (AUC) value and the
receiver operating characteristics (ROC) curve. The
average forecast accuracy over the 1000 iterations
was estimated at 0.74, with an average AUC
value of 0.82.
6 Tweets were gathered more Huge amounts of information likewise prompted the Twitter
Skuza et al (2015) than multi months’ choice to
time frame from second apply a decreased rendition of the Naive Bayes
January 2013 to 31st March calculation.
2013.
7 Network based Informal communities permit the impacted populace User behavior in
AltugAk ay et al. (2016) analysis, T F-IDF in light of Depression
score, k-means clustering perceptions got from client conduct in misery related Forum
informal communities.
8 Designed computational It is justifiable that there are different elements that Movie trailer
Yimin Ho u, et. al. framework composed could impact watchers like or aversion of the
(2016) Feature extraction, Feature trailer. To address inclination question. In view of
screening, Selection and mixed media highlights to figure out clients Like or
Classification Dislike of film trailer.
9 Suman, Gupta & Machine learning Relate the progression of stock cost Stock Twists
Sharma (2017)
10 Yuliyanti, Djatna & AI and Dictionary based Achievement level of the local area advancement Twitter
Sukoco. (2017) program
11 Hassan et al (2017) Machine learning Find the downturn level of an individual Twitter and
newsgroup
12 Karamollaoğlu, Lexicon-based Measure the insight or impacts of the peculiarities Twitter
Doğru, Dörterler,
Utku & Yıldız (2018)
13 Mansour. (2018) Lexicon-based Most client view ISISas a danger and dread Twitter
14 Facebook page
Poecze et al (2018) Machine learning Upgrade brand correspondence and grasping of YouTube
customer criticism Gamers
15 Martin-Domingo et al Machine learning Analyse airport service quality Twitter
(2019)
16 Nandal et al. (2020) Aspect level Machine In this paper Support Vector Machine (SVM) used Amazon
learning used in Aspect with diverse kernels like linear, RBF, linear kernel reviews

IJISRT24JUN1645 www.ijisrt.com 2150

Volume 9, Issue 6, June – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24JUN1645

level Sentiment analysis and polynomial

17 Adoma et al. (2020) Machine learning Worked ondifferent seven types of Emotion ISEAR
Detection like fear, Happy, sorrow, , irritation etc
with 79% Accuracy
18 Mukherjee et al. Deep Learning and Machine In this,Author apply different Machine learning Amazon
(2021) learning and Deep Learning models with reversal process Cellphone
and got Accuracy up to 95.30%. reviews
19 Kumar and Sachdeva Deep Learning and Machine Here author addressed Cyber victimization MyS- pace data
(2021) learning Identification using My Space dataset with 93.9% dataset
accuracy.
20 Liu et al. (2022) Machine learning In this paper Hybrid Sentiment Analysis model IEMO CAP
used for sentiments classification and find
multimodal reactions.
21 Mewada and Deep Learning and Machine BiLSTM used for features with probability with IMDB
Dewang(2023) learning 91.9% Accuracy but not find strong with handling
anonymous documents samples

IV. CONCLUSION [8]. Mansour, Samah. (2018) “Social Media Analysis of

User’s Responses to terrorism using sentiment
As per the studied of various review paper of analysis and text mining.” Procedia Computer
Classification Techniques and Approaches associated Science 140: 95–103.
with sentiment Analysis, We conclude that Survey [9]. Hassan, Anees Ul, Jamil Hussain, Musarrat
analysis, analyzes the classification techniques which Hussain, Muhammad Sadiq, and Sungyoung Lee.
present a good performance compared with existing (2017) “Sentiment Analysis of Social Networking
approaches and some classification techniques have some Sites (SNS) Data Using Machine Learning Approach
limitations. Combination of different techniques to beat the for the Measurement of Depression”, in International
drawbacks and advantages of increase the performance of Conference on Information and Communication
sentiments analysis. So we suggest that Hybrid Technology Convergence (ICTC), Jeju, South Korea:
techniques are more useful to find the appropriate IEEE.
results.To improve performance measurements in [10]. Karamollaoğlu, H., İ A. Doğru, M. Dörterler, A.
future, further effort will be required. Utku, and O. Yıldız. (2018, 20-23 Sept. 2018).
“Sentiment Analysis on Turkish SocialMedia Shares
REFERENCES through Lexicon Based Approach”, in the 2018 3rd
International Conference on Computer Science and
[1]. M.R. Saleh, M.T. Martín Valdivia, A. Engineering.
MontejoRáez, and L.A. Urena Lopez. [11]. Poecze, Flora, Claus Ebster, and Christine Strauss.
2011.Experiments with SVM to classify opinions in (2018) “Social Media Metrics and Sentiment
different domains, Expert Syst. Appl. 38, 14799- Analysis to Evaluate the Effectiveness of Social
14804 Media Posts.” Procedia Computer Science 130: 660-
[2]. Bing Liu.2012 Sentiment Analysis, Mining 666
opinions, Sentiments, and Emotions. [12]. Suman, N., P. K. Gupta, and P. Sharma. (2017, 11-
[3]. J. Bollen, H. Mao, X. Zeng, Twitter mood predicts 12 Dec. 2017). “Analysis of Stock Price Flow Based
the stock market, J. Comput. Sci. 2, 1-8. (2012) on Social Media Sentiments”, in the 2017
[4]. W. Medhat et al., Sentiment analysis algorithms and International Conference on Next Generation
applications: a survey, Ain Shams Eng. J. (2014). Computing and Information Systems (ICNGCIS).
[5]. Ravendra Ratan Singh Jandail., A proposed Novel [13]. AltugAkay, Andrei, and Bjorn – Erik Erlandsson
Approach for Sentiment Analysis and Opinion 2016. Assessing Antidepressants Using Intelligent
Mining, International Journal of UbiComp (IJU), Data Monitoring and Mining of Online Fora, IEEE
Vol.5, No.1/2, April 2014 Journal of Biomedical and Health Informatics, Vo l.
[6]. Yuliyanti, Siti, Djatna, Sukoco Taufik, and Heru. 20, No. 4.
(2017) “Sentiment Mining of Community [14]. Yimin Hou, Ting Xiao, Shu Zhang, Xi Jiang,
Development Program Evaluation Based on Social Xiang Li, Xintao Hu, JunweiHari, Lei Guo, L.
Media.” TELKOMNIKA (Telecommunication Stephen Miller, Richard Neupert, and Tianming
Computing Liu 2016. Predicting Movie Trailer Viewers
[7]. Martin-Domingo, Luis, Juan Carlos Martin, and Glen Like/Dislike via Learned Shot Editing Patterns,
Mandsberg. (2019) “Social Media as a Resource for IEEE Transactions on Affective Computing, Vo l.
Sentiment Analysis of Airport Service Quality 7, No. 1.
(ASQ).” Journal of Air Transport Management.

IJISRT24JUN1645 www.ijisrt.com 2151

Volume 9, Issue 6, June – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24JUN1645

[15]. Z. Jianqiang and G. Xiaolin 2017.Comparison [29]. Sneh Paliwal, Sunil Kumar Khatri and Mayank
research on text pre-processing methods on Twitter Sharma 2018 Sentiment Analysis and Prediction
sentiment analysis,'IEEE Access, vol. 5,2870_2879. Using Neural Networks, Proceedings of the
[16]. Y. Liu 2017.Social media tools as a learning International Conference on Inventive Research in
resource, J. Educ. Technol. Develop. Exchange, vol. Computing Applications (ICIRCA),
3, no. 1, 101_114. [30]. Mondher Bouazizi, Tomoaki Ohtsuki. (2017).A
[17]. T. Chen, R. Xu, Y. He, and X. Wang 2017. Pattern-Based Approach for Multi-Class Sentiment
Improving sentiment analysis via sentence type Analysis in Twitter IEEE Access, Vol.5, 20617 –
classification using BiLSTM -CRF and CNN,' Expert 20639.
Syst.Appl., vol. 72, 221_230 [31]. Wanxiang Che, Yanyan Zhao, Honglei Guo, Zhong
[18]. G.Vinodhini and R. Chandrasekaran 2017.A Su, Ting Liu.2015.Sentence Compression for
sampling based sentiment mining approach for e- Aspect-Based Sentiment Analysis,” IEEE/ACM
commerce applications, Inf. Process. Manage., vol. Transactions on Audio, Speech, and Language
53,no. 1(Jan): 223_236, Processing, Vol.23, No.12, 2111 – 2124.
[19]. G. Rao, W. Huang, Z. Feng, and Q. Cong 2018. [32]. H. Tang. S. Tan. X. Cheng2009. A survey on
LSTM with sentence representations for document- sentiment detection of reviews, Expert system
level sentiment classi_cation,'Neurocomputing, vol. application 36,10760-10773.
308(Sep): 49_57. [33]. Lu Lin, Jianxin Li, Richong Zhang, Weiren Yu and
[20]. M. Tubishat, N. Idris, and M. A. Abushariah 2018. Chenggen Sun2014. Opining mining and
Implicit aspect extraction in sentiment analysis: Sentiment analysis in social networks: A
Review, taxonomy, oppportunities, and open retweeting structure-aware approach, In
challenges,Inf. Process. Manage. vol. 54, no. 4(Jul): IEEE/ACM 7th international conference on utility
545_563. and cloud computing 890-895
[21]. B. Ghaddar and J. Naoum-Sawaya 2018.High [34]. El Rahman, Sahar A., Feddah Alhumaidi AlOtaibi,
dimensional data classification and feature selection and Wejdan Abdullah AlShehri. "Sentiment analysis
using support vector machines,'' Eur. J. Oper. of twitter data." 2019 international conference on
Res.,vol. 265, no. 3(March): 993_1004 computer and information sciences
[22]. P. Karagoz, B. Kama, M. Ozturk, I. H. Toroslu, and [35]. V.Ramanathan and T. Meyyappan, "Twitter Text
D. Canturk 2019.A frame-work for aspect based Mining for Sentiment Analysis on People’s
sentiment analysis on turkish informal texts,''J. Feedback about Oman Tourism", the 4th MEC
Intell. Inf. Syst., vol. 53, no. 3(Dec): 431_451 International Conference on Big Data and Smart
[23]. R. Priyantina, I. Teknologi Sepuluh Nopember, R. City (ICBDSC),2019,pp. 1-5.
Sarno, and I. TeknologiSepuluh Nopember 2019. [36]. Nandal N, Tanwar R, Pruthi J (2020) Machine
Sentiment analysis of hotel reviews using learning based aspect level sentiment analysis for
latentDirichlet allocation, semantic similarity and amazon products. Spat Inf Res 28(5):601–607
LSTM,'' Int. J. Intell. Eng. Syst., vol. 12, no. 4(Aug): [37]. Adoma AF, Henry N-M, Chen W (2020)
142_155 Comparative analyses of bert, roberta, distilbert, and
[24]. M. Al-Ayyoub, A. A. Khamaiseh, Y. Jararweh, and xlnet for text-based emotion recognition. In: 2020
M. N. Al-Kabi 2019.A comprehensive survey of 17th International Computer Conference on Wavelet
arabic sentiment analysis, Inf. Process.Manage., vol. Active Media Technology and Information
56, no. 2, (March): 320_342 Processing (ICCWAMTIP), IEEE, pp 117–121.
[25]. ]. L. Li, T.-T. Goh, and D. Jin 2020.How textual [38]. Mukherjee P, Badr Y, Doppalapudi S, Srinivasan
quality of online reviews affect classification SM, Sangwan RS, Sharma R (2021) Effect of
performance: A case of deep learning sentiment negation in sentences on sentiment analysis and
analysis, Neural Comput. Appl., vol. 32, no. 9(May): polarity detection. Procedia Comput Sci 185:370–
4387_4415 379
[26]. V. Bobicev 2016 “Text classification: The case of [39]. Kumar A, Sachdeva N (2021) A bi-gru with attention
multiple labels,” IEEE Int. Conf. Commun., vol. and capsnet hybrid model for cyberbullying detection
2(Aug): 39–42 on social media. World Wide Web 25(4):1537–1550
[27]. P. Chikersal, S. Poria, E. Cambria, A. Gelbukh, and [40]. Ding Y, Yu J, Liu B, Hu Y, Cui M, Wu Q (2022)
C.E. Siong 2015.Modelling Public Sentiment in Mukea: multimodal knowledge extraction and
Twitter : Using Linguistic Patterns to Enhance accumulation for knowledge-based visual question
Supervised Learning 49–65 answering. In: proceedings of the IEEE/CVF
[28]. Ayush Dalmia, Manish Gupta, Vasudeva Varma conference on computer vision and pattern
2015.Twitter Sentiment Analysis The good ,the bad recognition, pp. 5089–5098
and the neutral ,Association for Computation [41]. Mewada A, Dewang RK (2023) Sa-asba: a hybrid
Linguistic. model for aspect-based sentiment analysis using
synthetic attention in pre-trained language bert model
with extreme gradient boosting. J Super Comput
79(5):5516–5551

IJISRT24JUN1645 www.ijisrt.com 2152