Machine translation is a hot topic in our industry right now. As MT is getting better, translation clients are getting more excited about cost reduction, but many translators feel insecure, threatened by a potentially game-changing competitor. Undoubtedly, it’s a valid concern. I do have some of these fears, too, but I tend to think of MT as a timesaving and value-adding technology, not something to be scared of. First, I love innovation and discourage resistance to new things. Even though they might be scary at first, I know that it’s better to change now than lag behind for years. Second, I actually believe that just as CAT tools, MT can be a great assistant to translation pros. By embracing it, translators can increase productivity and reduce errors. Of course, I’m not talking about deceiving clients by selling raw or slightly edited MT output. What I mean is using MT results as a very rough draft at the very beginning of the translation process—the one that translators and editors are going to change heavily. What are some of the areas where MT can add value?
Generally, the shorter the segments, the better the results. For example, almost every translator had to translate a long list of country names at least once in his/her career. MT is perfect for such applications. It’s also useful for localizing software because it handles short user interface items and simple messages pretty well. But as segments get longer and more complicated (tags, line breaks, etc.), the quality of MT starts to dwindle.
Delivers Good Suggestions Quickly
Because MT relies on a huge translation memory and retrieves translations immediately, it occasionally comes up with better translations than a human translator would. Imagine coming across a phrase that you don’t translate very often. There’s no immediate translation in your mind. Yes, you’ll find the right translation, but it might take a minute or two. MT is sometimes able to give the right translation to you immediately, saving those precious minutes.
Free from Human Error
MT reduces the risk of mistranslations resulting from lack of concentration. For instance, it’s easy to make an error such as translating “from” as “form” or “contract” as “contact.” When a translator is tired in the evening or works under pressure of an urgent deadline, this kind of errors isn’t at all uncommon. As a machine, MT is flawless in this respect and can save us from those shameful blunders.
Handles Difficult Spelling Easily
MT is good at translating words with difficult spelling such as “methylmethacrylate.” Just last week, we were close to missing a spelling error in an English to Russian translation, catching it at the last minute. A translator made a typo in the word “teratology,” translating it as “tetratology.” Both the editor and proofreader missed the error initially. MT is free from such typos and can usually be trusted with translation of words with difficult spelling.
Caveat: MT is a good servant, but a bad master. What it originally produces as a whole is in no way final. Translators should use it very carefully. In fact, the next post will cover some of the pitfalls associated with using MT. So stay tuned and meanwhile, check out this post about my predictions about who’ll win: human translators or machines.
I know it’s a controversial topic that many readers have a strong opinion about. I’d love to hear yours.