Norms of Valence and Arousal for 2,076 Chinese 4-Character Words

Abstract

This study describes an annotated dataset through psycho-linguistic annotations in controlled environment on valence and arousal for a large lexicon of 2,076 Chinese 4-character words. The purpose for the annotation is to provide affect-linked knowledge to text which can be used in affective computing using NLP techniques. Analysis to the annotated data indicates that valence and arousal fit the classical U-shaped distribution. Most importantly, the annotated results indicate that the same 2-character word that appears in different 4-character words can indeed show distinct affective meanings which implies that the affective meaning of 4-character words may not be compositional to its component words. The study on this annotated list of 4-character words not only has significance at the intersection of cognitive neuroscience and social psychology, but also has great value as a resource for affective analysis in NLP applications.