Towards Semantic Aspect-Based Sentiment Analysis for Arabic Reviews

Towards Semantic Aspect-Based Sentiment Analysis for Arabic Reviews

Salima Behdenna (Department of Computer Science, Faculty of Exact and Applied Sciences, Laboratoire d'Informatique d'Oran (LIO), Université Oran1 Ahmed Ben Bella, Oran, Algeria), Fatiha Barigou (Department of Computer Science, Faculty of Exact and Applied Sciences, Laboratoire d'Informatique d'Oran (LIO), Université Oran1 Ahmed Ben Bella, Oran, Algeria) and Ghalem Belalem (Department of Computer Science, Faculty of Exact and Applied Sciences, Laboratoire d'Informatique d'Oran (LIO), Université Oran1 Ahmed Ben Bella, Oran, Algeria)
Copyright: © 2020 |Pages: 13
DOI: 10.4018/IJISSS.2020100101
OnDemand PDF Download:
Available
$37.50
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Sentiment analysis is a text mining discipline that aims to identify and extract subjective information. This growing field results in the emergence of three levels of granularity (document, sentence, and aspect). However, both the document and sentence levels do not find what exactly the opinion holder likes and dislikes. Furthermore, most research in this field deals with English texts, and very limited researches are undertaken on Arabic language. In this paper, the authors propose a semantic aspect-based sentiment analysis approach for Arabic reviews. This approach utilizes the semantic of description logics and linguistic rules in the identification of opinion targets and their polarity.
Article Preview
Top

Introduction

As a result of the rapidly increasing number of opinionated posts on different social media such as (blogs, Face book, discussion groups,…), analysis people’s opinion has gained considerable attention recently.

“Sentiment analysis, or opinion mining, is the field of study that analyzes people’s opinions, sentiments, evaluations, appraisals, attitudes, and emotions towards entities and their attributes” (Liu, 2012). The growing field resulted in the emergence of three levels of granularity: (i) document level, (ii) sentence level and (iii) aspect level:

  • 1.

    Document Sentiment Classification: The goal of document level sentiment analysis is to classify the sentiment expressed in the whole opinion document as positive or negative (Liu, 2012). This level provides an overall opinion of the document on a single entity (Behdenna, Barigou, & Belalem, 2016);

  • 2.

    Sentence Subjectivity and Sentiment Classification: The goal of this level is to classify the sentiment expressed in each sentence as positive, negative or neutral (Liu, 2012);

  • 3.

    Aspect-Based Sentiment Analysis: Both document level and sentence level do not find what exactly people liked and did not like. Aspect level yields very fine-grained sentiment information. The goal of this level of analysis is to discover and summarize people’s opinions expressed on entities and/or their aspects. The difficulty of Aspect-based Sentiment Analysis (ABSA) tasks is due to the following reasons: (i) the opinion can be expressed on entity explicitly or implicitly (Qiu, 2015). (ii) Many sentences without opinion words can also imply opinions (Liu, 2012). (iii) The difficulty in locating the opinion target; where Opinion targets is entity and their aspects about which opinions have been expressed (Qiu, Liu, Bu, & Chen, 2011).

The most commonly used approaches for sentiment analysis are: Machine Learning (ML) and Lexicon-based approaches. Machine Learning approaches require labeling a corpus in advance, and several supervised-based techniques are used (Zhang, Wang, Wu, & Huang, 2007). Lexicon-based approaches exploit a sentiment lexicon which is either built from existing dictionaries, or generated from the corpus (Oard, Elsayed, Wang, Wu, Zhang, Abels,, ... & Soergel, 2006). Sentiment analysis approaches is summarized in Figure 1 inspired from (Ibrahim, & Salim, 2013).

Figure 1.

Sentiment analysis approaches

IJISSS.2020100101.f01

On the other hand, most researches’ effort in Sentiment analysis deals with English texts and very limited researches are undertaken on Arabic language. Arabic is the official language of 22 Arab countries (Korayem, Aljadda, & Crandall, 2016), spoken by around 422 million people1, it is the fastest-growing language on the web (Korayem, Aljadda, & Crandall, 2016). Therefore, the need for a designing system for Arabic language is increasing.

The current paper focuses on aspect based sentiment analysis (ABSA) for the Arabic language. More specifically, this work considers two ABSA tasks: Extraction of the opinion target (implicit or explicit) and detecting their polarity. To this end, the authors propose to employ linguistic rules combined with ontology (they employ T-Box and A-Box to describe the ontology).

Complete Article List

Search this Journal:
Reset
Open Access Articles
Volume 14: 4 Issues (2022): 2 Released, 2 Forthcoming
Volume 13: 4 Issues (2021)
Volume 12: 4 Issues (2020)
Volume 11: 4 Issues (2019)
Volume 10: 4 Issues (2018)
Volume 9: 4 Issues (2017)
Volume 8: 4 Issues (2016)
Volume 7: 4 Issues (2015)
Volume 6: 4 Issues (2014)
Volume 5: 4 Issues (2013)
Volume 4: 4 Issues (2012)
Volume 3: 4 Issues (2011)
Volume 2: 4 Issues (2010)
Volume 1: 4 Issues (2009)
View Complete Journal Contents Listing