Predicting Emotion Labels for Chinese Microblog Texts
Zheng Yuan, Matthew Purver · 2014
Abstract. We describe an experiment into detecting emotions in texts on the Chinese microblog service Sina Weibo using distant supervision with various author-supplied conventional labels (emoticons and smilies). Existing word segmentation tools proved unreliable; better accuracy was achieved using char-acter-based features. Accuracy varied according to emotion and labelling con-vention: while smilies are used more often, emoticons are more reliable. Happi-ness is the most accurately predicted emotion (85.9%). This approach works well and achieves 80 % accuracies for "happy " and "fear", even though the per-formances for the seven emotion classes are quite different.