In the last blog post we introduced machine translation (MT), why it could be perceived as a threat to the professional translation industry and what its limitations are. In this blog we’ll look at some of those themes more closely and explore how the two can co-exist.
Just recently, Google’s Babel project hit the news. Details are short but it promises to be a technology that will allow two people to speak to each other over the phone in their native language while Google translates for them. Fans of The Hitchhikers Guide to the Galaxy will remember the Babel Fish, which could be inserted into the ear and used to understand any language in the world! This is what Google (and others) are reportedly going after and what an amazing futuristic technology it would be. It’s worth pointing out here that it sounds more of an interpreting tool, rather than a written translation tool, and these are two very different disciplines.
The sceptics would argue that the practicalities of such a tool would make it of limited use to most people. Would it be anywhere near accurate? How would it cater to accents and mumblings? Dialects? What kind of delay would there be? And so on. But I’m of the firm belief that it is only a matter of time before such wrinkles get ironed out. It might be 10, 30 or another 100 years, but that’s still just a matter of time. In the meantime, surely getting the gist of what someone is saying is good enough, considering we often have to speak with very rudimentary language skills when travelling abroad anyway.
So, is this a threat to the translation industry? Maybe more to the interpretation industry really. This is a technology that tries to ease instant understanding of what someone is saying. It is not designed to sell a product or attract visitors. Creative translation is a different matter entirely and this is where machines are still sorely lacking.
And it’s not just in the creative realm where machines might struggle. We recently had a case where a Czech translator had to edit a website translation that referred to his country as Eastern European, which may have caused offence and therefore damage to the company’s brand. The Czech republic may sometimes be included in this group of countries because the definition of “Eastern European” is so loose. From a purely geographical point of view, the Czech Republic is not in the east of Europe at all, it’s about as central as you can get. However, Eastern Europe is not usually used to describe countries’ locations. It is often used as a grouping for former communist countries that were part of the Eastern Bloc. It is not hard to see why residents of modern-day Czech Republic might object to being grouped under the Eastern European moniker. The point I’m making here is that cultural sensitivity is a particularly thorny issue and would you trust a machine to tread carefully through this tangle of thorns?
With these issues, it’s easy to see the tendency among many translators of distancing themselves from MT. Some of this is a defence mechanism to something that is perceived as a threat to the profession. But a lot of it is to do with professional pride, and justifiably so, when their own work is compared to that of MT. A good professional translation will beat an MT translation every time in terms of writing quality. The best an MT translation can really hope for is to be as good as a professional human translation, but never better, at least with current technology. Even far into the future, it’s hard to see machines as better at writing than humans, but of course that’s a subjective question anyway (what constitutes “better”?).
Despite some of these quality issues, we are seeing how MT can also enrich our lives, such as the Babel project. Machines will also always be much faster and efficient than humans at repetitive tasks and sheer processing of data. So, the reality is that MT and human translation are going to have to co-exist.
They already co-exist in a direct way where MT is used to generate preliminary translations which can then be edited by a professional linguist. In reality, it is often much better for the linguist to start straight from the source material, rather than clean up a mess, but MT is getting better and some may still prefer to start from a draft. Both humans and machines will probably settle into specific roles: MT for word lookups, portable enhanced dictionaries, getting the gist and perhaps simultaneous interpreting; humans for creativity, nuance, cultural issues and subjective mediation (“Both are correct, but which is more suitable?”).
But there is one other area where they can co-exist, and that is with guiding the future of the industry as a whole. Rather than distancing themselves from each other as competitors, MT developers and translators should work together to improve MT, draw up guidelines and codes of ethics, and to give a boost to the industry as a whole. There is a lot of expertise in the industry and MT developers should tap into it as much as possible.