Expert Articles - Language Tools
Machine Translation (MT)
The subject of machine translation -- automatic translation through a software program without any human assistance -- has been the subject of much controversy recently: is it reliable? Can it replace a human translator? Is it leading to lower quality translations? These are questions feeding one of the liveliest debates currently shaking the translation industry. But what exactly is MT?
It is important to make a distinction between the three types of MT, as each of them involves a different process: statistical, rule-based, and hybrid.
Statistical Machine Translation (SMT)
Statistical MT translates entire sentences using repositories, usually finding a correspondence between one word and another. They are, for example, Google Translate Moses, or Language Weaver. These MT tools might help you to understand the basic meaning of a text but they will never produce a very good translation. For example, translating “I want cake and not pie, please.” from English to French in Google Translate will give you “Je veux gâteau gâteau et non pas, s'il vous plaît.”, that is to say “I want cake cake and not, please.” Translating is definitely not a piece of cake!
Rule-based Machine Translation (RBMT)
Rule-based MT, as indicated by its name, is based on grammar rules. It will translate in the best grammatical way possible, using lexicons. Some examples are Systran, Prompt, or Pars. These software programs usually require very tedious proofreading, and as they need to be “fed” with lexicons, they really only become efficient after long term use.
Hybrid Machine Translation
The Hybrid MT is a mix of the SMT and the RBMT. It compromises between the two processes while integrating a translation memory, thus producing translations of a better quality and accuracy. Some examples are Asia Online and Apptek. These are the tools that most translation agencies use, since they are the best way to use the terminology most relevant to the clients’ needs.
Editing and Proofreading
Once the MT has finished its job, the translation is, of course, far from perfect. That’s when the human translator comes into play for the necessary editing part. The translator will proofread and rewrite all that has been translated by the machine, in order to make the translation sound natural. Some sentences will be so poorly translated that they will need to be entirely reworked.
Controversy
As mentioned previously, MT is the subject of much controversy: some linguists still doubt that this tool helps achieve a higher quality translation, or even gain any time. It is also difficult to determine who should take care of the post-translation editing procedure, a procedure that is not supposed to be the job of a translator.
In conclusion, it seems that the question of the efficiency of MT remains unanswered. It can clearly be a useful tool but still requires the proofreading skills of a professional linguist, who will be able to adapt the translation draft to make it sound natural to the target audience.
