An improved method for the feature extraction of Chinese text by combining rough set theory with automatic abstracting technology

The Rough Set Theory can reduce features of Chinese text effectively [1], but it is often encountered that the reduction will need a very long time in the case of a large number of training sets [2]. To solve the problem, this article proposes a method of associating Rough Set Theory with Automatic Abstracting Technology (AAT). Firstly, by calculating the weight of each node-it consists of the Self-Frequency, Tree Frequency, Concept Generalization Degree and Concept Selection Degree -in the Concept Hierarchy Tree [3] which based on Tongyici Cilin semantic dictionary [4] [5], it can determine theme concepts of Chinese Text. Secondly, it will extract the topic sentences [6] by calculating the importance of sentences [7]. Finally, it reduces features of these topic sentences again by IQR (Improved Quick Reduct Algorithm), and constructs the vector. Then from the whole information retrieval system perspective, it is clear that this method can save time for Automatic Abstracting and reduction. © Springer-Verlag Berlin Heidelberg 2012.