A non-parametric LDA-based induction method for sentiment analysis

Abstract

Sentiment analysis is the process of analyzing the characteristics of opinions, feelings, and emotions which are expressed in textual data. This paper presents a novel approach for generation of a lexical resource named PersianClues used for sentiment analysis in Persian language. Moreover, a novel unsupervised LDA-based sentiment analysis method called LDASA is proposed. In order to generate the PersianClues, at the first phase, an automatic translation approach is used to translate the existing English clues to Persian. Next, iterative refinement approach is used to correct the erroneous clues resulted from previous step. Then, topic-based polar sets are achieved from these clues and finally, each document is categorized into its related polarity using a classification algorithm. To evaluate this method, three resources about hotels, cell phones and digital cameras have been manually gathered from the e-shopping websites and the results of sentiment analysis on these resources are compared with a baseline named SVM-Unigrams. The experimental results demonstrate an improvement of 9% on average in polarity classification accuracy of the base system.