Discovering the Features in Opinion Mining via Domain Dependent and Domain Independent Relevance

Opinion feature mining is also known as aspect mining used to take out users opinions, and attitudes towards a specific product, services and their characteristics. The most of the existing approaches to opinion feature extraction on mining patterns is only by using a single review corpus. This paper presents the new method to discover the opinion features from online reviews by taking out the difference in opinion feature statistics across two different corpora, one domain specific corpus and another is domain independent corpus (i. e. the contrasting corpus). Domain relevance is the measure which is used to capture the disparity. The domain relevance characterizes the relevant term from the text collection. Firstly, the sentences are extracted from the reviews. Then the POS Tagger is applied to separate out the nouns, noun phrases and adjectives. Next the candidate features are extracted by applying the syntactic rules designed for Standard English. For every candidate feature, the Intrinsic Domain Relevance (IDR) and Extrinsic Domain Relevance (EDR) scores are calculated by using Domain dependent and domain independent corpus respectively. a The interval threshold approach, called as IEDR Criteria is applied to confirm the final Opinion Feature in which the candidate feature having IDR score greater than IDR threshold, and EDR scores less than EDR threshold is checked.