Construction and Application of Chinese Emotional Corpus

摘要

Text affective computing or sentiment analysis is an important research domain for natural language process, and it requires a large-scale emotion corpus, which can significantly support emotion classification and recognition. In this paper, some experiences and basic rules about constructing a Chinese emotional corpus are discussed, which include data collecting coverage, annotation system and quality criterion. Now there are 163,813 sentences labeled in our Chinese emotional corpus, and works like automatic emotion ontology learning, emotion transformation, gender analysis and emotion schema mining are available based on its statistic and labeled data. This effectively proves the valde of our Chinese emotional corpus.