Improved TF-IDF Feature Selection Method Based on Ontology Relative Degree
Weiyu Ma · Information Sciences · 2011
A method of improved feature extraction based on Ontology was proposed to compensate for the weakness of Traditional TF-IDF that Traditional TF-IDF does not consider the relation between the words.This method gets a set of candidate feature words which are the previous n words and a set of non-candidate feature words by Traditional TF-IDF,and gets a set of ontology associated concepts by the ontology relative degree;last,adjusts the weights of the feature keys by the ontology relative degree and the weights of ontology relative terms,and obtain the new results.The experimental results display that the new method improves the accuracy of feature extraction.