Using Synonym and Definition WordNet Semantic relations for implicit aspect identification in Sentiment Analysis
El Hannach Hajar, Mohammed Benkhalifa · 2019
Sentiment analysis (SA) is the process by which information can be extracted from customer reviews to be analyzed and categorized as positive, negative or neutral opinions according to three levels of SA: document, sentence and aspect levels. In aspect level, the aim is to determinate the target of sentiment or opinion represented in customer reviews, known as aspect term identification. In this paper, we address the aspect identification task involving implicit aspect implied by adjectives and verbs. Our approach considers the training data enhancement for Naïve Bayes (NB) classifier through the use of synonym and definition relations from WordNet (WN). In our method, we use the two semantic relations differently. NB classifier was used with five unique subsets of synonym and definition words and applied on two benchmark datasets, products reviews and restaurant. The obtained results show that: (1) The use of synonym and definition relations as represented in our work supports NB classifier and (2) Synonym and Definition relation must be appropriately combined the used to better support each implicit aspect identification task involving adjectives and verbs.