| Abstract | Past work on emotion processing has focused solely on detecting emotions, and ignored questions such as `who is feeling the emotion (the experiencer)?' and `towards whom is the emotion directed (the stimulus)?'. We automatically compile a large dataset of tweets pertaining to the 2012 US presidential elections, and annotate it not only for emotion but also for the experiencer and the stimulus. The annotations reveal that most of tweets express emotions of the tweeter, and only a few are indicative of the emotions of someone else. We then develop a classi er for detect- ing emotion that obtains an F-score of 55.86 on a 7-way classi cation task. Finally, we show how the stimulus iden- ti cation task can also be framed as a classi cation task, circumventing more complicated problems of detecting en- tity mentions and coreferences. Our classi er outperforms competitive baselines and obtains an F-score of 58.30. |
|---|
