A Study On The Chinese Text Summarization Method Based On Concept Lattice

With the popularization of computers, mobile phones, and the rapid development of modern information technology, there are a lot of new information added to the network and presented to the public in the form of electronic documents. How to obtain the information we need quickly and accurately has become an urgent problem. Obviously, the auto-summarization method is a practical way to solve these problems, but also can solve the problems of the small mobile device screen which is too small to read the large text information.We summarized and analyzed the study-situation of the auto-summarization method. And we proposed a method of text summarization based on concept lattice. The main contents can be summarized as follows:Firstly, this paper proposed a concept extraction method based on semantic similar degrees. We define the ―concept‖ as a word set which has the same meaning. Above all, we do the work about parting words, statistics frequency, and deleting the stopped words. Then, we calculate the weight of word using the information of frequency, word length, characteristic, and so on. After that, we calculate the semantic similarity of words which we have deleted the words having a small weight. And then, we can merge the similar words into a set called ―concept‖. Finally, we can export the first k ―concept‖ set according to their weight.Secondly, we select the ―concept‖ as the attribute and the sentences as the object to build a concept lattice which can express the information of a document. And then, to solve the problem about the large amount of computation, we reduce the attribute and the group of concept which is rare. And it is worthy of mentioning that the concept lattice we build has some study value in the following respects: compound words discovering, local topics discovering and similarity computing of sentences.Finally, we proposed a method of summarization using the above concept lattice. This method choose the minimum loss rate of concept as a measure, and export the summarization that has been selected by the optimization method. It can provide a set of sentences that has the minimum loss rate of concept.Experiments on the data sets provided by Fudan University show that the method we proposed is efficiency and accuracy, especially prominent in terms of the concept loss rate.