Implicit Sentiment Analysis of Chinese Texts based on Contextual Information and Knowledge Enhancement

Siqi Wang,Wenqiang Zhang,Zhenghui Cao,Haofen Wang

Abstract:The implicit sentiment analysis task makes the traditional explicit sentiment analysis methods no longer applicable due to the obscure and ambiguous expression of the sentiment text it focuses on and the lack of sentiment cue words containing obvious sentiment tendencies. In this paper, we propose a sentiment analysis method incorporating contextual information and external common sense knowledge for implicit sentiment analysis. We first propose a retrieval method to extract valid sentiment triples from a knowledge base and build an adaptive fusion of external common sense knowledge embedding layers to extend the semantic information of implicit sentiment expressions. Multipolar orthogonal attention mechanism is then used to learn embedding representations for implicit sentiment expressions, and an attention layer that incorporates context is introduced to mine and combine valid information in the context. Finally, the common sense knowledge information provided by the external knowledge base is fused with the contextual representation and the semantic information of the sentiment expression itself to achieve an effective extension of the implicit sentiment sentence representation and to perform sentiment label prediction. Experimental results on the SMP2019-ECISA dataset show that the method outperforms existing state-of-the-art methods and can fully exploit contextual information and external common sense knowledge to effectively improve the analysis of implicit sentiment expressions.

Linguistics,Computer Science