关键词: 树自动机推理算法, 结构化（半结构化）文档, 无秩树自动机, 信息抽取, KLH树

Abstract: This paper proposes an information extraction method based on an improved k-contextual tree automata inference algorithm.The key idea is to transform （semi-） structured documents into tree，creating unranked tree automata which can accept the tree and extract data according to the unranked tree automata state of acceptance and rejection，using an advanced k-contextual tree language，which is called KLH tree language.The method makes full use of the tree structure of the web document and combines the method based on web structure with grammar inference.Experimental results show that the approach with tree automata inference is favorable against some other approach in the learning time and extraction time.

Key words: tree automata inference algorithm, （semi-）structured documents, unranked tree automata, information extraction, KLH tree language