Based on the clustering technology in data mining, we aimed to establish a new schoolwork identifying mechanism. In order to let the normal answer can adapt to actual situations better, we first generalized the normal answer, and then calculated the similarity between every sample and normal answer, as well as similar degree between school works. Based on the similarity, we clustered all school work texts, and gave the one most similar to the standard answer the highest grade. Certainly, the number of clusters is appointed by teacher. Finally, we calculated the similarity between each cluster group and the standard answer, and in accordance with these calculation results, determined the final grade of the cluster groups.