Text Feature Extraction Based on the Extension of Topic Words and Fuzzy Set

Text feature extraction is one of the foundation of natural language processing, the traditional TF-IDF weight calculation method only consider characteristics of the frequency, but those feature items on different positions have different contributions to the text classification. By considering characteristics of the frequency, position and the mutual relations, an improved weight calculation method TF-IDF-Rel has been proposed based on the extension of topic words, and on the basis of this, embedding fuzzy set theory for the discretization of weight values. Experiment shows that this method is better than the traditional TF-IDF method of classification, and the recall rate and the accuracy rate have improved.