Using Bilingual Parallel Corpora in Translation Memory Systems

Automatic word alignment techniques commonly used in Translation Memory systems tend basically to work at single word level where there is a one to one correspondence between words in subsequences of the two languages. This, results in not being able to fully use subsentential repetitions like clauses, phrases and expressions. In this paper, using spaces between words, a search method named "space-based reduction search" is introduced. The main goal is to maximize the use of parallel corpus resources. We want to show that this search method can significantly enhance the chance of finding matches for subsequences of input sentences; hence applicable in a Sub-Sentential Translation Memory (SSTM) system without running automatic alignment tools.