Abstract

An important part of information-gathering behaviour has always been to find out what other people think and whether they have favourable (positive) or unfavourable (negative) opinions about the subject. This survey studies the role of negation in an opinion-oriented information-seeking system. We investigate the problem of determining the polarity of sentiments in movie reviews when negation words, such as not and hardly occur in the sentences. We examine how different negation scopes (window sizes) affect the classification accuracy. We used term frequencies to evaluate the discrimination capacity of our system with different window sizes. The results show that there is no significant difference in classification accuracy when different window sizes have been applied. However, negation detection helped to identify more opinion or sentiment carrying expressions. We conclude that traditional negation detection methods are inadequate for the task of sentiment analysis in this domain and that progress is to be made by exploiting information about how opinions are expressed implicitly.