Toward Context-aware Sentiment Analysis

Existing automated sentiment analysis mining methods either employ a static lexicon-based approach or a supervised machine learning approach. Nevertheless, the former method often fails to identify context-aware semantics of the opinion indicators, and the latter approach requires a large number of human labelled training examples. The main contribution of this paper is the development of a novel sentiment analysis method underpinned by context-aware text mining and probabilistic language modelling method to improve the effectiveness of sentiment analysis. Our initial experiments show that the proposed sentiment analysis method outperforms the purely lexicon-based method in terms of several benchmark measures. Our research opens the door to the development of more effective sentiment analysis method to perform business analytics.