Abstract:

As a new network language, emoticons have earned the favor of an increasing number of Microblog users. Emoticons in microblogs vividly represent blogger’s emotions. We first make a comprehensive analysis of emoticons in a large corpus of Chinese microblogs, including their usage, distribution and characteristics in emotion expression. Secondly, we manually select a list of emoticons that typically indicate six basic emotions as seeds. Based on the cooccurrence between a target emoticon and the seed emoticons in a large corpus, we establish sixdimensioned vectors for the target emoticon and apply them to emotion analysis. Experimental results on two data sets show that the emoticon vectors can effectively improve the precision of microblog emotion recognition.

Key words: emoticon;emotion vectors;statistical analysis;emotion analysis