An opinion detection algorithm based on online posts' relation

In recent years, opinion detection in posts has attracted more attentions, which can help us to understand people's reaction to the events effectively. In opinion detection, the representation of the posts' relations is important. Most of the existing researches utilized the vector space model and some similarity computing methods to describe the relations of the posts. But they ignored that the posts' relations are complex containing both the semantic part and the content part. In this paper, we first proposed a new hybrid post relations combining semantic similarity and content relation. Then we detected the opinions of online news posts with the networks of the hybrid posts' relations. The results show that our methods have better performance.