Main Article Content
Abstract
Opinion mining (also known as sentiment analysis) aims to analyze people’s opinions, sentiments, and attitudes facing entities such as products, services, and their attributes. Information retrieval is the process of extracting the information’s based on the occurrences of the terms in the document. We discuss about the method to identify features from online reviews by extracting the difference opinion feature statistics across two different large numbers of documents namely domain specific corpus and domain independent corpus. Defining a set of syntactic dependence rules, we extract the list of candidate opinion features from the domain review corpus. For each extracted candidate feature, we estimate a Intrinsic domain relevance, which represents the statistical association of the candidate to the given domain corpus. The Extrinsic domain relevance, which reflects the statistical relevance of the candidate to the domain independent corpus. The candidates with IDR scores exceeding a predefined intrinsic relevance threshold and EDR scores less than another extrinsic relevance threshold are confirmed as valid opinion features.