Deciphering Review Comments: Identifying Suggestions, Appreciations and Complaints

摘要

The problem of classifying sentences into various categories, arises frequently in text mining applications. One of the most important categorization of sentences observed in product reviews, movie reviews, blogs, customer feedbacks is - Suggestions, Appreciations and Complaints. We observed that the document classification techniques do not perform well for these three non-topical sentence classes. We propose to solve this problem using a supervised approach based on Dependency-based Word Subsequence Kernel and its variations. We compare the performance of our approach with the state-of-the-art short text classification techniques on 2 different datasets - Performance Appraisal comments and Product Reviews.