Domain Specific Emotion Lexicon Expansion

Abstract

Emotion Classification using lexicons has vast number of applications ranging from social media analysis to pervasive computing. Lexicons are usually hand-crafted and cost a lot of time and effort to generate. The major research challenge in this area is the creation of a generalized lexicon which serves well for every domain. This work focuses on automatic expansion of emotion lexicons to ease the process of domain adaption. Our proposed approach — CB-Lex — relies on a seed lexicon and an unlabeled corpus from the target domain. In experimental results, our expanded lexicons show an improvement of over 6% in F-Measure.