Implicit Aspect Based Sentiment Analysis for Customer Reviews

Abstract

This paper proposed a technique for sentiment analysis is the study that analyzes emotions, sentiments, and feeling based on the dictionary. There is a problem that the dictionary of sentiment does not contain sufficient words or field-specific words. Due to this issue, the meaning of the whole sentence can change into positive, negative, or neutral. It cannot express correctly, so the accuracy of text classification is decreased. In this paper, a dataset is used, which contains customer reviews, area-specific aspects, and emotion words. It explains the method that is manually annotated reviews, calculates the emotion scores to opinion words, and constructed the desired dictionary. Besides, the sentence layer is added in the latent dirichlet allocation (LDA) topic model. The proposed methods present an aspect based sentiment analysis (ABSA) that combines a sentiment dictionary and topic model. The experiment results show that the advanced method has the best performance on the dataset.