Polarity analysis of reviews based on the omission of asymmetric sentences

Resumen

In this paper, we present a novel approach to polarity analysis of product reviews which detects and removes sentences with the opposite polarity to that of the entire document (asymmetric sentences) as a previous step to identify positive and negative reviews. We postulate that asymmetric sentences are morpho-syntactically more complex than symmetric ones (sentences with the same polarity to that of the entire document) and that it is possible to improve the detection of the polarity orientation of reviews by removing asymmetric sentences from the text. To validate this hypothesis, we measured the syntactic complexity of both types of sentences in a multi-domain corpus of product reviews and contrasted three relevant data configurations based on inclusion and omission of asymmetric sentences from the reviews.