Novel relevance model for sentiment classification based on collision theory

Abstract

The performance of an Information Retrieval system is very much dependent on the effectiveness of the relevance model being used. Motivated by the concepts in Collision Theory in Physics, this paper proposes a novel approach of identifying relevance between two text objects. The role of positive and negative features is considered in designing the relevance measure based on the transitions in Collision Theory. For evaluating the measure, we have applied our relevance model on sentiment classification. © 2012 ICST Institute for Computer Science, Social Informatics and Telecommunications Engineering.