Clause-based approach to extracting problem phrases from user reviews of products

Abstract

This paper describes approaches to problem-phrase extraction from user reviews of products. The first step in problem extraction is to separate sentences with problems from all others. We propose two methods to problem extraction from such sentences: (i) a straightforward algorithm that does not split sentence into clauses and (ii) an improved clause-based algorithm.We claim that both approaches improve the classification performance compared to machine-learning algorithms.