Detecting Opinion Sentences Specific to Product Features in Customer Reviews using Typed Dependency Relations

Customer reviews contain opinions of the customers who purchased products and expressed opinions concerning their satisfactions and criticisms. Due to vast availability of product reviews in the web, it is extremely time-consuming and at times confusing for a new customer to manually analyze the reviews prior to buying a product. Reviews generally involve the presence of product feature specific factual information along with the opinion sentences depicting the pros and cons of a bought product. The unstructured format of the text reviews from most of the web review sources necessitates the automatic identification of opinion sentences from the customer reviews, and also the identification of explicitly visible and implicitly present product features associated with the opinion sentences. In this paper, a process has been described where typed dependency relations such as open clausal complements or adjectival complements have been utilized to identify opinion sentences specific to product features. The typed dependency relations in the identified opinion sentences are then used to associate a product feature to an opinion sentence with the help of the product feature associated frequent words extracted from a previously managed customer review corpus.