Detecting contextual word polarity using aspect based sentiment analysis and logistic regression

Sentiment analysis (SA) is a process done computationally for detecting opinion as well as determining their polarity. Context dependent opinion words remains as a challenge for SA since their polarity changes according to the context in which they are used. This work proposes a new approach for solving this problem using (aspect, opinion) pairs and logistic regression (LR) model. Syntactic rules are used for obtaining (aspect, opinion) pairs. LR model is employed as the classifier for classifying the sentiment words into positive class or negative class. In this work we have considered product review dataset and the evaluation results obtained showed an improved classification accuracy.