Corpus construction on polarity shifting in sentiment analysis

Abstract

Polarity shifting has been a challenge to automatic sentiment classification. In this paper, we create a corpus which consists of polarity-shifted sentences in the product reviews, where both the sentimental words and shifting trigger words are annotated. In particular, we group the polarity-shifted sentence structures into five main categories, i.e., negation, contrastive transition, modality, implication, and irrelevance Evaluation shows the statistics on the agreement of the annotation and the distribution of the five categories of polarity shifting is given. © 2013 Springer-Verlag.