Crowdsourcing is a new innovation in web-based translation that is growing in effectiveness and popularity.
While the antipathy towards machine translation in the translation services world has abated somewhat, the fact is you always need humans to make translation really work. Machines can help us, but they can’t replace us – and likely never will.
Language is simply too complex for computers to understand in any true way, and they evolve far too quickly for any computer system to keep up with. By the time you’re done performing the monumental programming tasks involved in training a computer to perform translations, the target language has developed a whole new set of slang terms and retired several dozen words that no one uses anymore!
The main problem is context: A computer can identify words and match them to a dictionary to come up with the correct word in the target language, and even do some limited grammar application. That’s why you can perform simple translations online these days. But computers like the awareness that comes with being able to sense context and construe meaning from the surrounding text. And that’s where a new concept in ‘crowdsourcing’ comes in.
Crowdsourcing for the Win
Crowdsourcing is the concept of letting a massive number of people work on a project. This can be small scale or large scale. On the small side, I can go to the Internet and post a question, like ‘What do people in Kenya say to each other when they need directions?’ and then sit back while hundreds of responses come in. Then I can collate the responses and choose the one that is suggested most.
On the larger scale, we have concepts in legal translation where volunteers spend a little of their time looking over and offering translations of text. They can work on whole pages or just a few sentences, and their work is placed into a database where it can be compared with other offerings. Over time, translation are certified by the ‘crowd’ and the entire text is translated, reviewed, and approved by the unseen ‘crowd.’
The strength of crowdsourcing is in numbers: The more people are involved, the better and more accurate the work is. Wikipedia is an excellent example of crowdsourced material: Yes, there may still be some mistakes in Wikipedia, but in general the crowd polices itself very well.
Crowdsourcing plus Technology
Now imagine we add a component of Translation Memory to our Crowdsourced Translation project. You visit a web page and a portion of it is automatically translated for you – the portions that are straightforward and simple, previously reviewed and confirmed – and all that’s left are the challenging and more difficult passages. The database of translation memory grows exponentially as people around the world contribute phrases to it, until eventually only the technical and difficult stuff needs working on.
Pretty powerful stuff. And you’ll note that with all this technology, the key remains the human beings involved.
Image courtesy thestaffingstream.com