Fine-grained sentiment analysis of reviews using shallow semantic information

There is a growing interest in sharing personal opinions on the Web, such as product reviews, economic analysis, political polls, etc. Existing research focuses on document-based approaches and documents are represented by bag-of-word. However, due to loss of contextual information, this representation fails to capture the associative information between an opinion and its corresponding target. Additionally, several researches focus on sentence-based approaches, which can effectively deal with an attribute-sentiment word pair within one sentence. However, those approaches are unable to process more than one attribute within one sentence. In this paper, we first present an improved sentiment word quantitative method to generate sentiment score for every word in sentiment lexicon. Additionally, we propose a novel identification approach of attribute-modifier-sentiment word triple using shallow semantic information. Experimental results show the feasibility and effectiveness of our approach.