Abstract:

We introduce a novel bigram topic model for analyzing the evolution of social emotions (BTMESE). Specifically, the proposed model can discern the semantic relationships among time, texts, and social emotions by creatively integrating wordorder information. In particular, the model could be applied to a wide range of application scenarios,such as emotion prediction and time prediction. The experiments based on real world datasets validate the model's effectiveness in terms of social emotion analysis.

Key words: wordorder;emotion analysis;social media;topic model