Text mining is nothing but the discovery of interesting knowledge in text documents. But there is a big challenging issue that how to guarantee the quality of discovered relevant features. And that are in the text documents for describing user preferences because of the large number of terms, patterns and noise. For text mining there are basically two types of approaches; one is term based approach and another is phrase based approach. But term based approach suffered with the problem of polysemy and synonymy. And phrase based approach suffered with low frequency occurrence.