Abstract-Mining the implicit knowledge in the electronic documents is a critical task in text analysis and data mining. To attain a knowledge-based view of the electronic documents, the clustering method based upon the topic cannot only be used, but also that based upon the extraction can be done. Therefore, a novel method for the clustering of the electronic documents, summarizing of the full text based on the extracted segments, and an evaluation using multi-measures for the importance to the document were presented. In the method, eighteen kinds of named entities and two kinds of syntactical phrases were extracted, and exploited for the text clustering. Then, a novel similarity equation was proposed for the calculation about the extractions. Meantime, three measures for the importance to the document were proposed, which provided a different view for the document's content, and recommended a prior checking for the users. Therefore, the method can improve the efficiency of the knowledge discovery, and enhance the management of the document on the large scale of document collection.