Use syntax phrases of different level to improve BoW
Phrases of different level in parse tree have different level of semantic abstract, and may function diversely in classification. This paper uses syntax phrases of different level to improve BoW representation. The result shows that level of phrases is valuable to capture the commonness of positive instances, and can better the discernment of positive instances. But it decreases the discernment of negative instances on the other side. And we also find BoW is sufficient as for negative instance recognition if there are enough positive instances.
syntax phrase parse tree text classification BoW
Ziqiang Li MingtianZhou
School of Computer Science and Engineering University of Electronic Science and Technology of China School of Computer Science and Engineering University of Electronic Science and Technology of China
国际会议
上海
英文
372-376
2010-06-22（万方平台首次上网日期，不代表论文的发表时间）