Opinion mining feature extraction using domain relevance

Abstract

The feature extraction of opinions from online user reviews is a task to identify on which features user is going to write a review. There are number of existing approaches for opinion feature identification but, they are extracting features from a single review corpus. These techniques ignore the nontrivial disparities in distribution of words of opinion features across two or more corpora. This proposed work discusses a novel method for opinion feature identification from online reviews by evaluation of frequencies in two corpora, one is domain-specific and other is domain-independent corpus. This disparity is measured using domain relevance. The first task of this proposed work is to extract candidate features in user reviews by applying a set of syntactic dependence rules. The second task is to measure intrinsic domain relevance and extrinsic domain relevance scores on the domain-independent and domain-dependent corpora, respectively. The third task is to extract candidate features that are less generic and more domain-specific, are then conformed as opinion features.