Research Of Single-document Summarization Technology Based On Sliding Window Extraction

Automatic summarization in searching information decreases the time of searching requisite information, increases the efficiency of knowledge-requirement . Therefore, the study of a automatic summarization techniques which can be propitious to all kinds of document has extremely important significance.Subject extraction is one of the key technologies in the summarization system. A good set of keywords can be a better reflects the central idea of the document, so as to extract information in the topic sentence to lay a good foundation. This paper first proposed sliding window-based keywords extraction algorithm, followed by the establishment of keywords based on the undirected graph, then an important degree of nodes to model and made words the weight variance and weight of keywords offset two evaluation indicators, to analyze the sliding window length on the keywords extracted impact.Theme sentence extraction is directly related to the quality level of automatic summarization, the collection of theme sentences is the final result of the automatic summarization system. Based on the extracted theme words, undirected graph based on theme words is extended, and the undirected graph based on the sentence is proposed. Sentence extraction problem is transformed to computing undirected graph node weights. Nodes, edges and edge weights in the undirected graph are determined in turn, and the node weights are computed finally. In order to determine the graph edge weights, sentences in the document are modeled by the Vector Space Model(VSM), and the relationship between sentences are clarified by the similarity. Finally, weights of sentences in the document are modeled by the weight model based on the similarity matrix, and the final output is obtained.Experiments show that the proposed automatic summarization techniques improves the recall rate and accuracy effectively.