Ontology-Based Approach to Enhance Explicit Aspect Extraction in Standard Arabic Reviews

Abstract

Currently, one of the most important and difficult areas of research is Arabic Sentiment analysis. Aspect extraction is the most important task in aspect-based sentiment analysis. In this work, we propose an aspect-based sentiment analysis approach for Arabic language. The basic idea of the proposed approach is to use the description logics to identify explicit aspects as well as their polarities. This approach was evaluated using the publicly available human annotated Arabic dataset (HAAD) based on Semantic Evaluation 2014 workshop (SemEval-2014) along with its baseline experiments conducted by HAAD providers. Experimental results showed that the proposed approach significantly improves the performance compared with baseline approach. The F-measure was improved by 19% and 55% for the aspect terms extraction and aspect term polarity tasks respectively.