It is an important problem how to choice the feature words which can express the topic exactly,and consequently reduce space dimension and time complexity.

We proposed a clustering computing based on feature words.

Different than stressing genealogical lines as is done by comparing feature words of a Pan-Min nature,the extraction of feature words pertaining to Hainan Min dialect is aimed at revealing its peculiarity.

This article uses correlative theory and researched date, whose representative characteristic words are from Binyang dialect, Tingzi Pinghua and Hengxian Pinghua, at the same time, com paring with characteristic words from Characteristic Words in Cantonese whose author is Zhang Shuangqing, to analyse the using situation.

This paper introduces the definition and its extract method of Hakka dialects characteristic words, and lists some examples,then stresses the analysis of modern Chinese words in it in order to show the close relation of Hakka dialects and the modern Chinese.

With the point of characteristic words, it’s very necessary and important to do the comparison.

To categorize documents hierarchically according to their topics,the thought of Fisher linear discriminant is utilized to get positive feature words and negative feature words in each category, and the algorithm of a hierarchical document categorization is given based on Fisher linear discriminant (HDCF).

To categorize documents hierarchically according to their topics,the thought of Fisher linear discriminant is utilized to get positive feature words and negative feature words in each category, and the algorithm of a hierarchical document categorization is given based on Fisher linear discriminant (HDCF).