A corpus-based study on Chinese sentiment parameters of Chinese sentiment discourse

Abstract

Most previous work on sentiment identification and annotation has focused on the identification and annotation of attitudes and targets, while less work has been done on other sentiment parameters. In this paper, we aim to discover different lexical, syntactic and semantic features of Chinese sentiment parameters based on Appraisal Theory. The data are from an annotated corpus of Chinese commentaries, analyzed using both qualitative and quantitative methods. We find that sentence-level sentiment production is the collaborative work of the core sentiment parameter (attitude) with other peripheral sentiment parameters (topic, source, field, process and degree of attitude). The distribution of sentiment parameters is also restricted by word classes, syntactic and semantic features and functions. This work not only offers a new analytic framework for Chinese sentiment analysis, but will improve the precision of sentence-level machine extraction of sentiment expressions in Chinese, with implications for possible extension to other languages.