Subsegment recall in Translation Memory — perceptions, expectations and reality

Abstract

Some CAT tool vendors describe their software as providing ‘subsegment’ matching, sometimes called ‘advanced leveraging.’ The descriptions can seem quite similar, but different tools in fact provide very different subsegment matching techniques and performance, with no typology available to distinguish implementations. This article first describes subsegment matching, then proposes just such a typology. The results of the first survey of translators to gauge their interpretations of vendor descriptions and their expectations of the software are presented and analysed in terms of the typology. A matrix of all CAT tools providing subsegment matching and available for trial (or free) use by translators is used to cross-reference their features with the typology. Finally, for the four CAT tools that provide the more advanced functionality, the first detailed analysis of their subsegment matching performance is presented, using an extensive series of tests. The overall findings show that interesting subsegment matching functionality is available, but that performance could be improved to meet translator expectations better. This in turn helps highlight the functionality to translators who may benefit from it, and the differences so as to help them make informed CAT tool purchase decisions.