Study On Chinese Text Automatic Summarization Based On Concept Extension And Integrated Evaluation Method

Along with rapid increase of scientific activities and their productions, scientific documents have been updating and improving quickly. Moreover, information retrieval has been the necessary means of collecting relative resources and their evaluations owing to Internet's booming. Consequently, automatic summarization which utilized computer and artificial intelligent technology, is easy to perform information retrieval and reuse, and has become the requirement of era.Single document summarization and multi-document summarization are researched and demonstrated respectively. The present work stresses to fulfill the latter based on concept extension and topic cluster in line with Chinese characteristics.Firstly, a comprehensive approach for single document summarization is discussed, and the evaluation method is the primary research object. The performance among different designed systems is compared, and the factors affecting the quality of summarization evaluation are also analyzed. All of methods have become the basis of the evaluation for multi-document summarization.Secondly, multi-document summarization is designed in relation to the above research rather than based on it. The research contents include sentence similarity computation, topic information identification, novelties' acquiring and summarization generation. Concept feature is expanded by altering the data structure of Hownet and using concept searching algorithm. By this process, problem of lack of semantic information of VSM are handled to a certain extent. The evaluating methods involve quality and content of presentation and information evaluation based on Q&A.At the same time, it is interesting to find out that the quality of evaluation on relationship and information is high, whereas the quality of other properties is not very satisfied in one summarization test. Therefore, according to the characteristic and virtues of all evaluating methods, a synthetic evaluation system based on integrated evaluation method is imported. The experiment results show that the research methods are effective.