User consultation questions are carefully analyzed and the new calculate of the text similarity model are pointed out—Abstract Knowledge Model. Using this model, the user's problem is decomposed into two parts: the Key-words and the Abstract Knowledge. Thus, the text similarity computing of user consult is transformed into Key-words and Abstract Knowledge similarity calculation. In addition, the paper has been improved the algorithm of the text similarity calculate in the past. Experimental results show that the proposed Abstract Knowledge Model plays a greatly role in text similarity calculation accuracy and the results closer to human judgment results.