Word Polarity Disambiguation Using Bayesian Model and Opinion-Level Features

Contextual polarity ambiguity is an important problem in sentiment analysis. Many opinion keywords carry varying polarities in different contexts, posing huge challenges for sentiment analysis research. Previous work on contextual polarity disambiguation makes use of term-level context, such as words and patterns, and resolves the polarity with a range of rule-based, statistics-based or machine learning methods. The major shortcoming of these methods lies in that the term-level features sometimes are ineffective in resolving the polarity. In this work, opinion-level context is explored, in which intra-opinion features and inter-opinion features are finely defined. To enable effective use of opinion-level features, the Bayesian model is adopted to resolve the polarity in a probabilistic manner. Experiments with the Opinmine corpus demonstrate that opinion-level features can make a significant contribution in word polarity disambiguation in four domains.