The knowledge-based information extraction approach has been proved effectively in the recognition and classification of Web texts. Manual creation of knowledge base is time consuming and error prone, even for a small application domain. For improving the automatic construction ability of knowledge base in the process of IE, this paper presents a web text extraction tool based on case similarity calculation (named WTE). WTE facilitates the process of constructing case knowledge base and strengthens the knowledge representation ability of single case by calculating case similarity and merging the cases which have similar semantic relationships. For validating the performance of WTE, experiments demonstrate that WTE provides higher recognition efficiency.