Automatic Emotion Annotation of Dream Diaries

This paper explores automatic annotation of dream re-ports. We first compile “emotion dictionaries ” from a thesaurus using Hall/Van de Castle emotion categories proposed for dream analysis. We capture the inherent ambiguity and polysemy of emotion words in word pro-files, which give the fuzzy membership score of a word on all five emotion categories. We then use the derived dictionaries to assign emotion categories to texts in so-called category profiles. We analyze several parameters of the technique. A comparison with the manual an-notation of texts from DreamBank indicate that this multi-faceted approach is promising. 1