Identify Sentiment-Objects from Chinese Sentences Based on Skip Chain Conditional Random Fields Model

Sentiment-objects Extraction aims to identify the targets of opinion described in sentiment sentence. Previous research fails to deal with the long-distance dependencies in Chinese sentences such as opinion targets repeated and echo of the different part of sentence. In this paper, we describe a probabilistic approach that incorporates the long-distance dependencies to identify opinion targets. The skip-chain Conditional Random Fields (CRFs) is used to model the long distance dependencies between sentiment sentences such as the repeated word and similar expression. Experiments show that our method outperforms linear-chain CRFs based method, and it is effective to identify opinion targets from Chinese sentences.