Mining construction rules of Chinese keyphrase based on rough set theory

Phrase conveys more information than word, and can better represent main topic of one article. Most of keywords we referred to are actually in form of phrases. The problem is that extraction of keyphrase lacks guidance of some general rules. By taking advantage of the ability of rough set theory on data generalization and knowledge reduction, the manually labeled keyphrase corpus which come from People's Daily was mined and some construction rules of Chinese keyphrase has been generated. These rules can be used for automatic keyword extraction, and can also help people manually label keyword. The experimental results are promising: the performance of keyword extraction improved greatly after importing these rules.