Emotion Category Mapping to Emotional Space by Cross-Corpus Emotion Labeling

The psychological classification of emotion has two main approaches. One is emotion category, in which emotions are classified into discrete and fundamental groups; the other is emotion dimension, in which emotions are characterized by multiple continuous scales. The cognitive classification of emotion by humans perceived from speech is not sufficiently established. Although there have been several studies on such classification, they did not discuss it deeply. Moreover, the relationship between emotion category and emotion dimension perceived from speech is not well studied. Aiming to establish common emotion labels for emotional speech, this study elucidated the relationship between the emotion category and the emotion dimension perceived by speech by conducting an experiment of cross-corpus emotion labeling with two different Japanese dialogue corpora (Online Gaming Voice Chat Corpus with Emotional Label (OGVC) and Utsunomiya University Spoken Dialogue Database for Paralinguistic Information Studies (UUDB)). A likelihood ratio test was conducted to assess the independency of one emotion category from the others in three-dimensional emotional space. This experiment revealed that many emotion categories exhibited independency from the other emotion categories. Only the neutral states did not exhibit independency from the three emotions of sadness, disgust, and surprise.