6 Limitations of Computer Translation

February 25th, 2013

There’s nothing we can do about it: machine translation is here to stay (on some level, at least).

6 Limitations of Computer Translation | One Hour Translation

Still Not Ready for Prime Time

Those nightmares are misplaced, of course, because machine translation still has too many limitations to be practical for any professional-level translation work of any complexity. Yes, we’ve gotten further than some thought possible not so long ago – the fact that you can visit a web page and get workable translations of simple sentence in a wide variety of languages is remarkable, and a tribute to the impatient genius of man – but there is so much distance between our current state and true human-free translation, I have not yet felt the need to speed up my retirement plans.

Six Limitations

There are a few areas where machine translation’s limitations still far outweigh its advantages:

Speed. Computers are getting faster all the time. When you surf to Google Translate and type in a sentence, the translation comes immediately! And yet ... you lose more time than you gain when you have to go over every output from the computer and fix its mistakes.

Subject Matter. The undeniable fact is, computers are not aware and do not ‘know’ anything in the strictest sense of the term. They are unaware of context and incapable to applying it.

Accuracy. Computers are blunt instruments in the legal translation world; you pretty much get the same performance all the time. I can vary the time and energy I spend on something depending on the needs of my clients. That flexibility does not exist with machines right now.

Vocabulary. While computers can be built to have immense dictionaries and, of course, perfect recall, what you’ll find is that the translations produced by them will start to sound the same, as the same words are used over and over again. When I’m doing a lot of work for a client in a single field or text, I try to vary things a bit while still maintaining accuracy.

Volume. The only time machine translation becomes a true option is when volume and not quality is the goal, but the bar for volume is very high. Unless you have immense amounts of text that must be translated by tomorrow, a human is still going to do a better job.

Expense. While it might seem like all the translation tools are free, those are really just the toys offered as loss leaders by Google and Microsoft. The professional machine translation tools are quite expensive – and yet carry all of the limitations just discussed.

You might also like: