Cognitive Detection of Multiple Discrete Emotions from Chinese Online Reviews

Sentiment analysis is a hot topic in a couple of years. Emotion, an affective state expressed by human cognitive process, is widely embedded in user-generated content (UGC). Traditional research mainly focused on polarity and paid less attention to the nature of emotion that elicited from an underlying cognitive structure with multiple discrete dimensions. Informed by the OCC emotion model with hierarchical cognitive structure, we firstly detect and extract discrete emotions from online reviews of JD.com, one of the most famous electronic product marketplaces in China. Secondly, we propose an improved OCC-OR emotion model and select six prominent discrete emotions elicited from customer online behaviors and conditions. Thirdly, we validate the proposed emotion model into natural language processing and machine learning techniques. The performances of multiple discrete emotion detection in distributed Chinese semantic representation and classifiers prove the effects of OCC-OR emotion model, which can better demonstrate consumer cognitions in online shopping market. Our findings build up a solid theoretical foundation in sentiment analysis with a practical implication for future researchers.