A Survey on Sentiment Analysis, Classification and Applications

Size: px

Start display at page:

Download "A Survey on Sentiment Analysis, Classification and Applications"

Dorthy Wheeler
5 years ago
Views:

Volume 119 No. 10 2018, 1203-1211 ISSN: 1311-8080 (printed version); ISSN: 1314-3395 (on-line version) url: http://www.ijpam.eu ijpam.

1 Volume 119 No , ISSN: (printed version); ISSN: (on-line version) url: ijpam.eu A Survey on Sentiment Analysis, Classification and Applications * Syed Saood Zia 1, Sana Fatima 2, IdrisMala 3, M. Sadiq Ali Khan 4, M. Naseem 5, Bhagwan Das 6 1* Department of Computer Engineering, Sir Syed University of Engineering and Technology, Karachi, Pakistan 2 Department of Computer Science, Sir Syed University of Engineering and Technology, Karachi, Pakistan 3 Department of Electrical Engineering, Usman Institute of Technology, Karachi, Pakistan 4,5 Department of Computer Science, UBIT, Karachi University, Karachi, Pakistan 6 Department of Electronic Engineering, Quiad-e-Awam University of Engineering, Science and Technology (QUEST), Pakistan szia@ssuet.edu.pk, imala@uit.edu, msakhan@uok.edu.pk, mnaseem105@gmail.com Abstract Sentiment analysis or opinion mining is the computational study of people s opinions, appraisals, attitudes, and emotions toward entities, individuals, issues, events, topics and their attributes. Sentiment analysis plays a vital role in the research area of text mining. The aim behind the sentiment classification is to analyze the core (subjective) information in the text and then categorizes into one of the three categories positive, negative and neutral. The purpose of this paper is to analyze different techniques for sentiment classification that assign a polarity to an opinionated text. This paper summarized the sentiment levels for classification, some widely used algorithms, possible challenges and application areas of sentiment classification. Keywords: Sentiment Analysis, social Media, Blogs, Forum, Polarity. 1. Introduction With the frenziedprogress of social media (i.e., different reviews, forum discussions, blogs and social networks) on the World Wide Web, individuals and organizations are increasingly using public opinions in these media for their decision making [1]. Social media not only plays an important role in connecting people with the world but it becomes an essential entity to helps people to express their emotions or feeling about any topic or opinion on the globe. Sentiment Analysis is an application of natural language processing that analyzes people s opinion towards different products or entities on social media and then classifies into three categories positive, negative and neutral [2]. Sentiment analysis is also called subjectivity analysis. It focuses on the main

2 subjective part in the text that lead the phrase or sentence in good or bad direction [3]. Opinion mining or opinion extraction is the study of opinions comes from different corners of the world. Sentiment classification follows four steps shown in figure 1. In the first step, collect different contents given by many users (reviews or feedback) from social media web sites. After that, apply preprocessing steps to clean the irrelevant datathat are not necessary for sentiment classification. After the preprocessing step, analyze the linguistic features in user generated text by using part of speech (POS) approach so that required information is identified [4].Then sentiment classification is performed by using machine learning mechanisms in order to determine the polarity of the text. User Feedback Data Cleaning Sentiment Analysis Assign Polarity Figure 1: Process of Sentiment Classification Most of the research exists on sentiment analysis for user opinion data, which mainly judges the polarities of user reviews [5]. In these studies, sentiment analysis is often conducted at one of the three levels: the document level, sentence level, or attribute level [6]. In relation to sentiment analysis, the literature survey done indicates two types of techniques including machine learning and semantic orientation. In addition to that, the natural language processing techniques (NLP) is used in this area, especially in the document sentiment detection [7]. This paper is organized as follows: Section 2 presents the sentiment analysis and their levels. Section 3 introduces the approaches for sentiment analysis. Section 4 presents some challenges in this area and section 5 identified the benefit of sentiment analysis. Last section concludes our study and discusses some future directions for research. 2. Level of Sentiment Analysis Sentiment analysis has been explored mainly at three levels: Document Level Analysis Sentence Level Analysis Entity and Aspect level Analysis 2.1.Document Level Analysis The task in the document level sentiment analysis is to classify whether a whole opinion document asserts a positive or negative sentiment [8, 9]. In this level, each document declare opinions on a single entity and it is not applicable in those situations where the documents which evaluate or compare multiple entities. 2.2.Sentence Level Analysis In this level, the task based on sentences and determines either each sentence conveyed a positive, negative, or neutral opinion. This level of analysis is not suitable where the sentence structure is complex [10]

3 2.3.Entity and Aspect Level Analysis Entity and Aspect level analysis are also called phrase level sentiment analysis. The above two mentioned level do not identify exactly what the people liked or did not like. Aspect level sentiment analysis performs finer-grained analysis. It is based on feature-based opinion mining that identifies the features values of the opinion and then summarize the results of that opinion [11]. 3. Approaches for Sentiment Analysis In general, the common approaches used in sentiment analysis are lexicon based approach and machine learning approach. 3.1.Lexicon based approach The Lexicon based approach aims to finding the lexicon containing the opinion and then analyzes it by either using the dictionary based approach or the corpus based approach [12, 13] Dictionary based approach The Dictionary based approach aims to finds the morphemes that contain subjective meaning in the text and then matches them from the words listed in the dictionary, but it has disadvantage that it is unable to find the domain or context specific opinion Corpus based approach The Corpus based approach matches lexicon opinion with the list of opinion words that identifies correct domain or context of specific opinion. Acquiring of the opinionated lexicon is done in two steps. In the first step, identified the word that contain opinion in the corpus and in second step, polarity is assigned to opinionated word. 3.2.Machine Learning Approach Machine learning investigates how computers can learn (or improve their performance) based on data. In this approach, how computer program can learn to recognize complex pattern and make intelligent decisions based on data [14]. It mainly comprises of supervised learning and unsupervised learning approach. In supervised learning approach, we mainly focus on classification of data while in unsupervised approach, we focus on clustering. Mostly the opinion are classified either positive, negative or neutral. On the basis of data classification, we have identified some supervised learning algorithms that are commonly used in sentiment analysis and opinion mining Naïve Bays Classifier (NB): One of the simplest and widely used supervised learning approach is Naïve Bays. It works on bags of words strategy i.e. a class C is assigned to document d is based on the allocation of the words in that document [15, 17]. The function of the NB classifier is as follows. It assumes the features presents in the document are independent on each other so probability of the document d belongs to class Cj can be calculated by the summation of all probabilities of all independent features in document d as mention below:

4 P X C i = P(x k C i ) n k=1 (1) = P x 1 C i ) P x 2 C i ).. P x n C i ) We can easily estimate the probabilities P(X 1 C i ), P(x 2 C i ),, P(x n C i ) from the training data. Bayes classifier model produces fastest and accurate results as well as very robust to irrelevant features. The only drawback of this classifier is that it assumes all features in the given dataset are independent on each other Support Vector Machine Classifier (SVM): Dataset based on text is ideally suited for this classifier because of disperse nature of the words in the text, there may be some features in the data set that are irrelevant but they tend to exist together in a way to organize into easily separable classes [16, 17]. In this technique a straight line curve is created between the data items that will classify them into multiple classes. If there are X items, W is the input feature vector then straight line curve can be determined as P = W. X + b (2) Where W is the weight factor and b is the scaling factor and X is the training tuples of data. SVM is independent on features gap and very fast in training as well as testing. It provides very high accuracy if given text belongs to one class only. The drawback of this technique is that if multi class items exist in the opinionated word the gap around the curve becomes reduces which in result, decrease the accuracy to correctly classifying those words Maximum Entropy The maximum entropy classifier is the best at solving sentiment classification problems [17]. Unlike NB, it does not assume that the features are independent of each other. This model uses search search-based mechanism to find out the uncertainty if the data is not clear. In other words we should select the model p* with Maximum Entropy as: p = arg max p x p y x log p(y x) x,y (3) Where P(x) is the experimental distribution of x in the training dataset. This classifier can be used to solve a large variety of text classification problems. The major disadvantage of this technique is, if the categories are unbalanced then result may be affected as it often guesses the largest category Decision Tree

5 This classifier describes the sentence structure in a form of a tree; each tree includes a root node, branches, and leaf nodes. Each internal node represents an attribute, each branch denotes the outcome of a test, and each leaf node holds a class label. it is very simple and powerful technique for multiple variable analysis [18]. Decision tree resolve problem in two iterative steps, at first it calculates Entropy which cleans the impurity in the data set by using following formula: E X = c i=1 Pi log 2 Pi (4) Where P i is the probabilities of class i and it calculate as the proportion of class i in the set. In second step information gain is calculated that tells us how important a given attribute of the feature vectors is Gain T, X = Entropy T Entropy (T, X) (5) Where T is the total entropy of parent root and X is the entropy of child (leaf) node. The above two steps are repeated continuously until the anomalies in the data set are removed. This classifier is robust and performs well with large data in short time. The drawback of this technique is that if there are too many nodes present in the tree that are correlated to each other, the complexity becomes so high and it is very difficult to classify text into a particular category because of this greedy structure. 4. Challenges Some of the major challenges faced in the area of sentiment analysis. Sentiment analysis is based on subjective classification so if there are more individuals involve in the research, the outcome may be the difference in opinion [19]. The biggest challenge faced in sentiment analysis is the domain specific nature of opinionated words. It may give very good performance in one domain, but at the same time it performs very poor in some other domain [20]. Social media websites facilitate user to provide feedback in any textual pattern this will create problem in determination of opinionated word in the text [21]. Since every human being has a different nature so it is very hard to correctly classify user provided input belongs to a particular entity. 5. Benefits of Sentiment Analysis In this section, we will expound some benefits of sentiment analysis in the real world environment. Sentiment analysis helps in business intelligence by analyzing the customer s reviews regarding a product. Sentiment analysis helps in the detection of spam on social media through automatic detection of the spam words on different forums, s and blogs. Sentiment analysis is very useful in monitoring response of people regarding political issues on social media by analyzing the arrogant words on different internet resources

6 6. Conclusion Sentiment analysis is the computational study of subjective word in the text which provides user s opinion towards entities o social media websites. In this paper, various techniques for sentiment classification are discussed. Each of them has certain advantages and disadvantages over other classifiers. Sentiment analysis plays a vital role in the research area of text mining which helps in order to retrieve some useful information of interest. Nowadays, it becomes the growing field of research. Progress is continue in order to resolve the challenges not only in English language but in other native languages as well. 6. References [5] C. E. Larsen, R. Trip, and C.R. Johnson, Methods for procedures related to the electrophysiology of the heart. U.S. Patent 5,529,067, Jun 25 [1] Liu, Bing, and Lei Zhang. "A survey of opinion mining and sentiment analysis." In Mining text data, pp Springer US, [2] Serrano-Guerrero, Jesus, Jose A. Olivas, Francisco P. Romero, and Enrique Herrera-Viedma. "Sentiment analysis: A review and comparative analysis of web services." Information Sciences 311 (2015): [3] Cambria, Erik, BjörnSchuller, Yunqing Xia, and Catherine Havasi. "New avenues in opinion mining and sentiment analysis." IEEE Intelligent Systems 28, no. 2 (2013): [4] Swami, Amogh, Ajit Mete, SurajBhosle, Nikhil Nimbalkar, and Sonali Kale. "Feature Extraction and Refinement For Opinion Mining." (2017). [5] Yadollahi, Ali, AmenehGholipourShahraki, and Osmar R. Zaiane. "Current State of Text Sentiment Analysis from Opinion to Emotion Mining." ACM Computing Surveys (CSUR) 50, no. 2 (2017): 25. [6] Soleymani, Mohammad, David Garcia, Brendan Jou, BjörnSchuller, Shih-Fu Chang, and MajaPantic. "A survey of multimodal sentiment analysis." Image and Vision Computing 65 (2017): [7] Rosenthal, Sara, NouraFarra, and PreslavNakov. "SemEval-2017 task 4: Sentiment analysis in Twitter." In Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017), pp [8] Turney, Peter D. "Thumbs up or thumbs down?: semantic orientation applied to unsupervised classification of reviews." In Proceedings of the 40th annual meeting on association for computational linguistics, pp Association for Computational Linguistics, [9] Pang, Bo, Lillian Lee, and ShivakumarVaithyanathan. "Thumbs up?: sentiment classification using machine learning techniques." In Proceedings of the ACL-02 conference on Empirical methods in natural language processing-volume 10, pp Association for Computational Linguistics, [10] Liu, Bing. "Many Facets of Sentiment Analysis." In A Practical Guide to Sentiment Analysis, pp Springer International Publishing, [11] Manek, Asha S., P. DeepaShenoy, M. Chandra Mohan, and K. R. Venugopal. "Aspect term extraction for sentiment analysis in large movie reviews using Gini Index feature selection method and SVM classifier." World Wide Web 20, no. 2 (2017): [12] Taboada, Maite, Julian Brooke, Milan Tofiloski, Kimberly Voll, and Manfred Stede. "Lexiconbased methods for sentiment analysis." Computational linguistics 37, no. 2 (2011): [13] Feldman, Ronen. "Techniques and applications for sentiment analysis." Communications of the ACM 56, no. 4 (2013): [14] Han, Jiawei, Jian Pei, and MichelineKamber. Data mining: concepts and techniques. Elsevier,

[15] By, Done. "Document Classification Method Based on Contents Using an Improved Multinomial Naïve Bayes Model diss., PhD ".

" Journal of Informetrics 3, no. 2 (2009): 143-157. [17] Go, Alec, RichaBhayani, and Lei Huang. "Twitter sentiment classification using distant supervision." CS224N Project Report, Stanford 1, no.

" IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews) 42, no. 3 (2012): 291-312. [19] Liu, Bing. "Sentiment Analysis and Subjectivity.

7 [15] By, Done. "Document Classification Method Based on Contents Using an Improved Multinomial Naïve Bayes Model diss., PhD ".ی سح تنمنجذوعتماددحلدودینفیابرطزینصتةقیافلوةقیثاانتسإادتحمیلوختساباھتایماد Middle East University, [16] Prabowo, Rudy, and Mike Thelwall. "Sentiment analysis: A combined approach." Journal of Informetrics 3, no. 2 (2009): [17] Go, Alec, RichaBhayani, and Lei Huang. "Twitter sentiment classification using distant supervision." CS224N Project Report, Stanford 1, no (2009): 12. [18] Barros, Rodrigo Coelho, Márcio Porto Basgalupp, Andre CPLF De Carvalho, and Alex A. Freitas. "A survey of evolutionary algorithms for decision-tree induction." IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews) 42, no. 3 (2012): [19] Liu, Bing. "Sentiment Analysis and Subjectivity." Handbook of natural language processing 2 (2010): [20] Medhat, Walaa, Ahmed Hassan, and HodaKorashy. "Sentiment analysis algorithms and applications: A survey." Ain Shams Engineering Journal 5, no. 4 (2014): [21] Pang, Bo, and Lillian Lee. "Opinion mining and sentiment analysis." Foundations and Trends in Information Retrieval 2, no. 1 2 (2008): Authors Syed Saood Zia Syed Saood Zia obtained his BSc (Hons) in Computing & Information Systems from London Metropolitan University in the year of He has completed his Master of Computer Science from KASBIT in 2004 and MS in Mobile Computing & Information Systems from Hamdard University in He has recently completed his PhD in the field of Information Technology in He has more than seventeen years of teaching experience in different academic institutions. He is working as Assistant Professor in Computer Engineering Department at SSUET. He is a member of British computer society. His research area is knowledge based systems, Information retrieval, Big Data Analytics, Sentiment Analysis, Opinion Mining and Data Mining. Sana Fatima Sana Fatima obtained her Bachelor of Engineering (B.E.) in Computer & Information System from NED University of Technology with first division. Recently, she has completed his Master of Computer Science & Information Technology from NED University of Technology with first division. She has more than five years of teaching experience in different academic institutions. She is working as Jr. Lecturer in Computer Science Department at SSUET. She is a member of Pakistan Engineering Council. Her research area is decision support systems, knowledge based systems, sentiment analysis and data mining. Idris Mala Idris Mala obtained his BE in Electrical Engineering from NED University in the year of He has completed his MS in Information Technology from Hamdard University in He has completed his ME in Telecommunication from Hamdard University in He has recently completed his PhD in the field of Information Technology in His research area is Fuzzy Logic,

Neural Networks, Advanced Database Management System, Advanced Algorithms and Big Data Analytics. M. Sadiq Ali Khan M.Sadiq Ali Khan received his Ph.

Since 2003 he is serving Computer Science Department, University of Karachi as an Assistant Professor.

He is the member of CSI, PEC, IEEE and NSP. He is currently Wise Chair IEEE Computer Society Karachi Section. Muhammad Naseem M.

8 Neural Networks, Advanced Database Management System, Advanced Algorithms and Big Data Analytics. M. Sadiq Ali Khan M.Sadiq Ali Khan received his Ph.D Degree from KU in 2011 and his BS & MS Degree in Computer Engineering from SSUET in 1998 and 2003 respectively. Since 2003 he is serving Computer Science Department, University of Karachi as an Assistant Professor. He has about 18 years of teaching and research experience and his research areas includes Data Communication & Networks, Network Security, Cryptography issues and Security in Wireless Networks. He is the member of CSI, PEC, IEEE and NSP. He is currently Wise Chair IEEE Computer Society Karachi Section. Muhammad Naseem M. Naseem obtained his BS in Computer Engineering from SSUET and MS in Computer Engineering with specialization in computer networks from SSUET. He is perusing his Ph.D. in Computer Science from UoK. He is working in the Department of Computer Engineering at SSUET as an Assistant Professor. His research spans various fields including Image Processing, Computer Networks, VLSI, Cryptography algorithm. Bhagwan Das Bhagwan Das received his Ph.D Degree from University Tun Hussein Onn Malaysia (UTHM), Malaysia in Oct He has obtained his BS in Electronic Engineering from MUET and MS in Electronic Engineering from QUEST. He is working as Lecture in Electronic Engineering Department QUEST.He has published more than 40 articles in international journals and in conferences these are indexed in SCOPUS, ISI, EIcompendex. He has presented his research work in Denmark, Sweden, Spain, Singapore, and Vietnam.His area of research includes optical system design, optical signal processing, and energy efficient optical communication

9 1211

10 1212

Techniques for Sentiment Analysis survey

I J C T A, 9(41), 2016, pp. 355-360 International Science Press ISSN: 0974-5572 Techniques for Sentiment Analysis survey Anu Sharma* and Savleen Kaur** ABSTRACT A Sentiment analysis is a technique to analyze