Machine translation (MT) is fast and cheap, but if you’ve ever used Google Translate or any other translation engine to translate a page of text, you’ll know that while it’s good by the standards of a few years ago, it’s a long way from being usable for translating critically important content (don’t, for heaven’s sake, let it loose on legal documents). When it comes to idiomatic usage and nuance, humans still rule—and given the significance of context in the translation process, it’s hard to see how this will change any time soon.
So, why does Smartling offer MT integration?
The answer is that MT is a viable option for some site owners, in some circumstances. Smartling client Vintage Parts, for example, have found it valuable in reaching customers in a total of 22 languages—something they could never realistically have achieved using any other method. (They later post-edited MT with human help – we’ll get to that in a minute.)
If you’re mulling over using MT (over the other options), here are some circumstances where you might consider it:
- Speed is a priority.
- No other method can match the ability of MT to handle hundreds of pages of text almost instantly.
- You can use professional or volunteer translators/editors to clean up the results afterwards. (Vintage Parts used this approach with professional translators post-editing the completed MT.)
- Price is a priority.
- MT on its own is the cheapest option—cost is currently $20 per million characters – about 200,000 words. Do be aware, though, that most users will need to take into account the cost of checking and correcting the outcome.
- Volunteer translation, which you might think is competitive on cost, really is not, as volunteers require considerable management.
- Accuracy is NOT a priority.
- There may be circumstances (e.g. translating large quantities of archived material that you still want to make available but is not front and center on your site) where MT may be better than no translation at all.
- You have extensive and repetitive lists to translate (e.g., a parts inventory).
- The absence of complete sentences and the predominance of repetitive text elements make lists natural candidates for MT.
- You have a crowd/professionals who can be let loose on the MT to clean it up.
- MT can be a good first step. But there’s really no substitution for human translation, and based on what the needs are, post-edited machine translation (PEMT) could be an even better second step. (This hybrid approach is a trend among Smartling users.)
MT is just one tool in the box, and a rather crude one at that—something for roughing out your work rather than for finishing it. There’s no cut and dried answer to the question of whether it’s for you. Any one or none of the circumstances I’ve laid out above may be a sufficient rationale for using it. Remember, though, that it’s not an all-or-nothing decision. You can—and probably should—use MT only on specific kinds of content. You can—and probably should—experiment with it first before using it extensively. But in the right circumstances, its combination of speed and low price can be impressively persuasive.