Firstly , the paper give the definition of model selection Agent . Based on the definition , the paper dis2 cusses the structure and the functions of the component . Next , the paper makes a research on the method how to re2 alize model selection Agent . During the process of realizing model seletion Agent , the corpus approach in linguistics is used. By building the corpus based on model selection case , the paper bring forward a new method of knowledge representation of model selection. On the basis of this new method , it gives a main algoristhm of the Agent mission processor.