An Unsupervised Sentiment Information Identification Approach

Abstract

Existing research focuses on document-based sentiment analysis and documents are represented by the bag-of-words model. However, due to the loss of contextual information, this representation fails to capture the associative information between an opinion and its corresponding target. Additionally, several researchers focus on sentence-based approaches, which can effectively extract an aspect-sentiment word pair within one sentence. Nevertheless, their approaches can only deal with one aspect within one sentence and miss the identification of sentiment modifier. In order to solve these problems, this paper proposes a novel identification approach of aspect-modifier-sentiment word triple using shallow semantic information. Experimental results show that our approach is feasible and effective.