Text fine classification method

Abstract

The invention provides a text fine classification method, which belongs to the technical field of computer natural language processing or pattern recognition and solves the problems of low fine category division accuracy of the existing text classification method on short files. According to the method, firstly, two stages of classifiers are built according to an existing training sample, wherein each stage of classifier comprises an independent sensitive dictionary; then, operations such as term segmentation, special pause term removal and synonym network mapping are executed on the training samples of the classifiers, and the preprocessing of term vectors is realized; next, the feature selection is carried out according to the importance difference of the term vectors, and the sensitive dictionary of the classifiers is built; then, a KNN (k-Nearest Neighbor) algorithm is used for calculating a fine classification result of a target file; and finally, the classification result is evaluated and feed back, the sensitive dictionary is dynamically optimized, and the classification accuracy is further improved. Experiments prove that when the text fine classification method provided by the invention is adopted for by the short files, the fine classification accuracy can be obviously improved.

Description

technical field [0001] The invention belongs to the technical field of computer natural language processing or pattern recognition, and in particular relates to a text fine classification method, which can improve the fine classification accuracy of short documents. Background technique [0002] Text classification generally includes the process of text expression, classifier selection and training, classification result evaluation and feedback, and text expression can be subdivided into text preprocessing, indexing and statistics, feature extraction and other steps. [0003] Traditional text classification methods usually classify long documents with obvious differences between categories, such as web content classification (sports, news, finance and military, etc.). However, in some specific fields, such as short document classification such as automatic classification of police information for public security and sentiment analysis of Weibo, the gap between categories is ...

AI Technical Summary

Problems solved by technology

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a text fine classification method to solve the problem of low accuracy of short document fine classification and significantly improve the accuracy of fine classification

Images