From Computer Assisted Translation to Human Assisted Translation.
Fiorenza Mileto · 2014
Automated translation systems have been developed over the past 50 years and in the latest years (thanks to the evolution from rule-based machine translation to statistical machine translation) they are penetrating the translation industry in a very massive way. It seems that they are ready to revolution the world of translation.Finally corpora (after years of use and misuse) found a good place to be used in a fruitful way: statistical machine translation needs huge quantities of words to work properly and to return an understandable sentence.At this stage of development, machine translation and assisted translation are working together not only because the post-editing process is performed mainly inside CAT tools, but also because translation memories are more and more used as corpora for specialized translation and specific subjects. Unfortunately, translation memories were not always created and used in a proper way in the translation industry: they are often “dirty” and badly managed because they are fed and used by so many translation specialists in the various stages of translation projects, they are exchanged among so many different tools with different default settings often unknown to creators and users, and, most of all, they are mixed and shared but rarely cleaned.From a CAT tool point of view, translation memories created for a specific subject are to be considered an asset, but a dirty translation memory is not a good starting point for machine translation: if you have to tell to the machine how to translate and you provide an inconsistent resource, the output of machine translation can hardly be good.Machine translation aims at reducing the efforts and the intervention on the translation side, maybe substituting the translator or reducing the traditional competences requested to the translator. Is it really so ?The initial effort to clean the linguistic data on which a machine translation engine is based is considerable and expensive. It requires all the traditional professional abilities and competencies that a translator acquires after years of studies and experience.After the revolution introduced by CAT tools, translation industry is preparing to face a new revolution. Universities may play a pivotal role in this revolution: they may represent a bridge between linguistic data available on the market and the traditional linguistic competencies required to prepare them for machine translation, teaching students how to clean and maintain TMs, how to post-edit a translation generated with machine translation and how to help linguists and developers of machine translation engines to improve them.Maybe machine translation is only an intermediate step between computer assisted translation and human assisted translation…