With a wide use of item bank system and the increment of items in item bank system, how to avoid duplicate items becomes an important research topic. This paper first gets text with vectors with TF-IDF formula through the algorithm based on vector space mode(VSM) theory. Then, it gets the similarity of items by using cosine theory, which is used for the comparison with the threshold value initialized to get similarity checking resulting. Based on the existing item bank system, the experiment with this algorithm shows that the exact rate of 94% is gained, which is a good result compared with expert checking.