A Translator's View about Translation Memory and Machine Translation Integration

Abstract

Translation memory (TM) systems have become a key technology to help translators in dealing with a steadily growing demand for translation, and areknown to have a positive effect on productivity, quality, and cost of translation. However, certain restrictions of this technology prompt developers ofcomputer assisted translation (CAT) tools to improve and expand the functionality of TM systems. Therefore, machine translation (MT) is beingprogressively integrated into TM tools. The paper aims at analysing potential advantages of TM and MT integration, discusses the effects of MT on awork routine of professional translators, and presents opinions of translators towards the integration of MT into the process of translation.

There are two ways of integrating MT into TM systems: MT systems are included into CAT software packages or translators are provided with apossibility to port their TM tools to commercial or free access on-line MT providers; according to the second scenario, TM systems are complementedwith statistical, example-based or rule-based MT technologies.

The poor quality of MT output made a lot of professional translators reluctant to use MT; however, the data-driven MT approach is making a difference.There are studies that compare translation speed when post-editing MT suggestions, reviewing repetitions and correcting fuzzy matches from a TMdatabase. Results demonstrate that the difference is not so significant and the MT output when post-edited by a human translator can also reach thedesirable level of the quality. Thus, technological competence and post-editing skills are being emphasised as the most significant skills for a professionaltranslator.

Although there are disagreements among translators towards the integration of MT into their workflow, potential productivity gains derived from theintegration of TM and MT technologies and changing requirements of the translation market with regard to speed and cost of translation are likely totrigger further developments in the automation of the process of translation.

DOI: http://dx.doi.org/10.5755/j01.sal.0.23.5559