The only detailed description of a proposal for solving the partitioning problem has been reported by Maruyama and Watanabe (1992). They describe an algorithm for determining the best partitioning (``cover'') of a source language sentence in an environment which, though not strictly EBMT, is closely related to EBMT. The input to their algorithm is a) a source language sentence and b) a set of pairs consisting of a substring of the input sentence and the translation of this substring into a target language. The output of their algorithm is, essentially, a string of source language substrings which completely covers the entire input sentence. The task which we set out to perform is more general than theirs along each possible dimension of comparison. Our assumptions have consistently been less constraining than those used by Maruyama and Watanabe. To name just a few, in addition to the operations carried out by Maruyama and Watanabe's system, we also a) determine the set of substrings that best match the input -- instead of stipulating the prior existence of the set of best matches; b) allow a fuzzy, incomplete match between input and corpus and c) do not proceed from the assumption (as do Maruyama and Watanabe) that each substring is relatively short. In a later report we will analyze these differences in greater detail.