Research Of Single-document Summarization Based On Semantics

With the development of the Internet and search engine, people have grown more concern to get more valuable information from Internet efficiently. Document summarization technology becomes a useful tool to solve this problem. Many universities and international enterprises create a variety of automatic summarization applications.In china, document summarization has a short history. Main research is based on statistical methods, such as the summarization based on Vector Space Model. Because of the undevelopment of linguistics research and lack of the various corpus and dictionary based on semantic, we focus on the single-document summarization based on semantic.Single-document summarization goals to create a compressed summary while retaining the main mean of the original document. Many approaches use statistics and machine learning techniques to extract sentences from a document. In this paper, we proposed two new single-document summarization frameworks based on semantics. Firstly we research the word similarity and sentence similarity computation based on hownet. Then we propose two methods, which are hownet based modified K-Medoids and symmetric non-negative matrix factorization auto summarization frameworks. Experimental results demonstrate the improvement of the summary quality by using our two proposed frameworks.