A Context Analytical Method Basing on Text Structure

摘要

In this paper the research techniques of complex network are introduced into the complement of missing data in text and a new method of text mining is put forward basing on the text structure of large-scale texts. First the GRE word net is constructed by using lots of relative articles specially for experiment, then the static characters of this network are analyzed and the context relationships of words are obtained in it according to the community discovery algorithm of complex network, next an complement algorithm is designed to judge whether it is the right complement words by following relationships among these words. In the experiment, we take the examination questions of GRE as test set and use this method to do the sentence completions in verbal sections, the result demonstrates the availability of this text analyzing method which focuses on topology information of network. It can not only apply to the imputation of missing data, but also the complement of full sentence after skeleton’s forming in machine dialogs.