The LSI model described in the previous section does not specify a particular implementation strategy. The abstract model can be broken into two phases: a preprocessing phase where the term-document space is built, and the query phase where queries are projected into that space and a similarity measure is applied to find terms and documents near the query. Although both phases of the LSI model are equally important, the preprocessing phase is a one-time cost. It is independent of the number of queries posed to the database in the second phase. Since the user must interact with the query phase, the functionality and execution efficiency of that phase is very important. Here, adequate software to perform the preprocessing phase is assumed to exist. Instead, the query phase is considered.