AdaTextTiling: A New Adaptive Method To Text Segment Base On TextTiling

With computers gradually popular in our daily lives,Internet information degree is increasing constantly.And people often use can bring huge Internet data resources,which contain the extremely high value.The text data is a main Internet data resources,and text mining is to excavate valuable information from the text data resources.As an important branch of text mining,text segmentation has a very important role in the text mining of information.Text segmentation refers to the whole text is composed of multiple sub-topics,then can use some methods to divide the whole text into multiple segments by sub-topics.There are a lot of text segmentation algorithms,Text Tiling algorithm is a very classic text segmentation algorithm.This paper is mainly to improve the classic TextTiling algorithm,and then put for-ward an advanced AdaTextTiling algorithm.At first,this paper is to analyze TextTiling algorithm,and then optimize it.The main point is to adjust the length of the text window when calculating text similarity of every potential point,because this paper argues that optimal text window length of every potential point is not fixed.And this paper also improves the computational efficiency with LDA topic model.Finally through the exper-iment,this paper found AdaTextTiling algorithm performance is significantly superior to TextTiling algorithm,which shows the effectiveness of AdaTextTiling algorithm.