According to the discussion of Independent Component Analysis(ICA),a fast fixed-point extract algorithm based on the kurtosis is developed,where the iterative steps get fewer,usually 3 to 15.

Using techniques of full text index and fuzzy binary searching,based on the full text index of the dictionary,this paper introduces two functions of fuzzy dictionary lookup and fast inputting words in a machine aided translation system These functions make of more conuenient to look up and input word

Based on a probe of procedures of realizing ATC,some proposals are presented for improving automatic categorizing,including building word-extracting dictionary with reductional method,selecting features with voting method,hierarchical text categorization with step-by-step categorization algorithm,and building categorization algorithm by integrating statistics wit.

Using the principles of rotated Keywords for reference and in combination with the relevant statistical models,this article compresses and optimizes the original word-extracting dictionary in both positive and negative directions so as to achieve the goal of reducing the number of dimensions and accurately expressing the themes.