Abstract: Latent Semantic Analysis (LSA) offers computational methods for extracting and representing the meaning of words as underlying dimensions of a large text corpus. This paper presents an automatic approach of using LSA to measure item similarity with the goal of identifying enemy relationship in item banks.