Logistically Supervised Aspect Category Detection Using Data Co-occurrence for Sentiment Analysis

Abstract

Word of mouth (WoM) always had an impact on consumer decision making. Before making any purchase decision, family and friends are asked for advice and recommendations. With the advent of Web, wlectronic WoM (EWoM) came into being. One of the important forms of EWoM is product and service reviews. These reviews contain large amount of information, and so, reading these reviews one at a time is a tedious and time-consuming task. Therefore, summarization of reviews is desirable. This can be done by identifying the aspects in the reviews. Aspects are the specific topic that is being referred. The proposed approach is a hybrid of three methods for aspect category detection. The first method is based on co-occurrence between the indicators and categories, and the second method is based on logistic regression, used to predict the probability of a dependent variable based on one or more features. The third one based on Wu-Palmer similarity measure. Experimental results show a good performance of the proposed hybrid approach.