作者: Laurence Vidrascu , Vered Aharonson , Dino Seppi , Thurid Vogt , Laurence Devillers
DOI:
关键词:
摘要: In this paper, we report on classification results for emotional user states (4 classes, German database of children interacting with a pet robot). Starting 5 emotion labels per word, obtained chunks different degrees prototypicality. Six sites computed acoustic and linguistic features independently from each other. A total 4232 were pooled together grouped into 10 low level descriptor types. For these groups separately all taken together, using Support Vector Machines are reported 150 the highest individual Information Gain Ratio, scale With both features, relative improvement up to 27.6%, going higher Index Terms: emotion, prototypes, feature types, automatic