Improving sentiment analysis in Arabic: A combined approach

Sentiment analysis SA is concerned with determining users’ opinions from text reviews. However, mining these opinions from massive amounts of unstructured and lengthy documents is a challenging task. In this paper, we propose methods that extract valuable opinions and values from online movie review...

Full description

Saved in:
Bibliographic Details
Main Authors: Belgacem Brahimi, Mohamed Touahria, Abdelkamel Tari
Format: Article
Language:English
Published: Springer 2021-12-01
Series:Journal of King Saud University: Computer and Information Sciences
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1319157819303283
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849323449603850240
author Belgacem Brahimi
Mohamed Touahria
Abdelkamel Tari
author_facet Belgacem Brahimi
Mohamed Touahria
Abdelkamel Tari
author_sort Belgacem Brahimi
collection DOAJ
description Sentiment analysis SA is concerned with determining users’ opinions from text reviews. However, mining these opinions from massive amounts of unstructured and lengthy documents is a challenging task. In this paper, we propose methods that extract valuable opinions and values from online movie reviews to improve SA in Arabic. First, we propose a method that explores the role of n-gram and skip-n-gram models in opinion classification. Second, we study a method that exploits subjective words such as adjectives and nouns by applying Part-Of Speech tagging. Both of the methods are combined with a feature reduction technique to enhance SA results. Third, we present a method that seeks to extract relevant opinions such as review summaries and conclusion opinions. Then, a combined approach is proposed to augment opinion classification results. Forth, we introduce a method for analyzing customers’ opinions by determining factors impacting their attitudes based on the costumer value model. Experimental results conducted on two datasets prove that our proposed methods are effective and provide better scores than baseline sentiment classifiers. The best obtained classification results reached 96% in F-Measure. These results indicate also that the aesthetic factor is the most influent factor in Arabic movie reviews.
format Article
id doaj-art-0433758789d94bee92da2dbf51a12e3d
institution Kabale University
issn 1319-1578
language English
publishDate 2021-12-01
publisher Springer
record_format Article
series Journal of King Saud University: Computer and Information Sciences
spelling doaj-art-0433758789d94bee92da2dbf51a12e3d2025-08-20T03:49:03ZengSpringerJournal of King Saud University: Computer and Information Sciences1319-15782021-12-0133101242125010.1016/j.jksuci.2019.07.011Improving sentiment analysis in Arabic: A combined approachBelgacem Brahimi0Mohamed Touahria1Abdelkamel Tari2Department of Informatics, Faculty of Exact Sciences, University of Bejaia, 06000 Bejaia, Algeria; Corresponding author.Department of Computer Science, University of Setif1, 19000 Setif, AlgeriaLIMED Laboratory, Faculty of Exact Sciences, University of Bejaia, 06000 Bejaia, AlgeriaSentiment analysis SA is concerned with determining users’ opinions from text reviews. However, mining these opinions from massive amounts of unstructured and lengthy documents is a challenging task. In this paper, we propose methods that extract valuable opinions and values from online movie reviews to improve SA in Arabic. First, we propose a method that explores the role of n-gram and skip-n-gram models in opinion classification. Second, we study a method that exploits subjective words such as adjectives and nouns by applying Part-Of Speech tagging. Both of the methods are combined with a feature reduction technique to enhance SA results. Third, we present a method that seeks to extract relevant opinions such as review summaries and conclusion opinions. Then, a combined approach is proposed to augment opinion classification results. Forth, we introduce a method for analyzing customers’ opinions by determining factors impacting their attitudes based on the costumer value model. Experimental results conducted on two datasets prove that our proposed methods are effective and provide better scores than baseline sentiment classifiers. The best obtained classification results reached 96% in F-Measure. These results indicate also that the aesthetic factor is the most influent factor in Arabic movie reviews.http://www.sciencedirect.com/science/article/pii/S1319157819303283Text miningOpinion miningSentiment classificationReview extractionCombined approach
spellingShingle Belgacem Brahimi
Mohamed Touahria
Abdelkamel Tari
Improving sentiment analysis in Arabic: A combined approach
Journal of King Saud University: Computer and Information Sciences
Text mining
Opinion mining
Sentiment classification
Review extraction
Combined approach
title Improving sentiment analysis in Arabic: A combined approach
title_full Improving sentiment analysis in Arabic: A combined approach
title_fullStr Improving sentiment analysis in Arabic: A combined approach
title_full_unstemmed Improving sentiment analysis in Arabic: A combined approach
title_short Improving sentiment analysis in Arabic: A combined approach
title_sort improving sentiment analysis in arabic a combined approach
topic Text mining
Opinion mining
Sentiment classification
Review extraction
Combined approach
url http://www.sciencedirect.com/science/article/pii/S1319157819303283
work_keys_str_mv AT belgacembrahimi improvingsentimentanalysisinarabicacombinedapproach
AT mohamedtouahria improvingsentimentanalysisinarabicacombinedapproach
AT abdelkameltari improvingsentimentanalysisinarabicacombinedapproach