科研之友
微信
新浪微博
Facebook
分享链接
摘要
Identifying comparative sentences and types is an important part of finegrained opinion mining. In the area of computational linguistics, there has been no research on Uyghur comparative sentences. Therefore, we deeply analyze the characteristics of Uyghur comparative sentences, and propose a method which combines bidirectional Class Sequential Rules mining (Bi-CSRs) with rules. The Bi-CSRs can effectively decrease the number of useless CSRs and the rules can improve the experimental performance. For the type identification, we use an ordered keyword strategy to identify types. The experimental results demonstrate the effectiveness of the method we have put forward in this paper.
