Translation Memory: Virtues and Vices
Translation memory (TM) equipment, the most widely used toolkits in the localization of digital information at HQ-translate agency, enable the linguistic transition and cross-border compliance of electronic content (e-content) for local markets. The idea behind TM systems is to store in a computer system the original e-content and the translation that has been produced by human translators; the stored translated version of the source text has been broken down into manageable bits, generally one sentence long. Today the most popular CAT tools: TRADOS, Déjà vu, Wordfast. The priviledges of using translation memory systems are fairly obvious: they increase the translator’s productivity and better translation quality by providing that terminology and sentences are used consistently within and across translation works. Users in business and cross-border agencies submit a 25–60% rise in work throughput. Yet, it must be stated that the use of TM systems may also have negative effects on translation quality. One of the major discouraging things of TM systems is that they usually perform at sentence level. That’s why, there is a serious danger that the translator will focus too much on isolated sentences, possibly disregarding the contexts in which the sentences are built-in. Moreover, the matching algorithms of TM systems are based on very plain formal criteria, such as the similarity of character strings. That’s why, the human translator’s notion of the grade of similarity between a segment to be translated and a part retrieved from the storage base may differ considerably from the grade of similarity calculated by the TM system. This may result in situations wherein exact matches result in wrong translations, or one translation of a fuzzy match requires little or no adjustment but another fuzzy match with the same similarity grade is not useful at all (for a discussion on the aspects of evaluating the retrieval mechanisms of memory systems, see Expert Advisory Group on Language Engineering Standards (1996), Whyman and Somers (1999), and Reinke (2000a, 2004). Despite the negative sides, it should be noted that TM systems generally build into the translation performance comparatively smoothly. These CATs leave human translators in control of the real translation process, while free them from routine work and supporting translation as a creative act whenever the adaptation resourcefulness of a human translator is required. For more remarks, visit us at: HQ-translate company