Chinese Document Summarization

The modern society is full of information. Along with computer' s popularization and the Internet' s development, massive electronic information appears to people every day. How to learn quickly the needed information from the magnanimous documents has become an important question. It is almost impossible for people to select what they aim to search by reading all the available documents. Therefore, an automatic information compression tool to abstract or condense the massive information becomes inevitably a must.Automatic abstracting is an important research direction in natural language processing. The purpose is to explore the mechanism of acquiring and abstracting information from texts and to develop the programs that can automatically write abstracts, therefore to improve the efficiency of information retrieval and information spread. The automatic abstracting has the following characteristics: (1) it ought to be able to automatically extract the original text' s main thoughts or the central contents. (2) Its returned digest should be summary, objective, understandable, and readable. (3) It should be suitable for any subject. Automatic abstracting is different from information extraction in that it does not prescribe the target' s characteristic in advance while the latter does and returns only the related information.Automatic abstracting aims to pick up the digest from the primitive literature through computer procedure. The digest should exactly reflect the literature' s major idea. In automatic abstracting, four methods are usually applied: Automatic extraction, understanding-based abstracting, information extraction, and structure-based abstracting.In this thesis, the Chinese abstracting system takes the method of automatically extracting the sentences from the original text to generate...