Notes from Action Week for Global Information Sharing 09Posted on Sep 30 2009 by George Weyman. Filed under: Social translation, Technology
The idea of bringing together translators with the people who craft and manage web-based translation tools seems on one level obvious.
Translators need good software to manage and expand their work; translation software companies need translators to tell them what works and what doesn’t.
On another level though, putting translators in a room with the people who are perhaps most responsible for establishing the machine as the first draft translator of choice for the masses (and thereby putting basic translation of texts in reach of billions of web users) is like sitting a high-end portrait artist in a room of budget camera makers.
In fact, one of the loudest messages to come out of Action Week for Global Information Sharing in Limerick last week was that translators need to see MT, and the Translation Memories that underpin it, as an unprecedented opportunity.
Translation memories have a huge potential in increasing the volume of translations translators can do. And with more and more translation jobs going as the web expands both audiences and content, translators would be crazy not to use them.
Kirti Vashee (@kvashee) even said that TMs could enable translators to do in three hours what would have previously taken them five.
But come on, now. How many translators actually know what TMs are, let alone use them? Hands up!
I think the truth is many are not aware of the value of TMs, a trend which is probably also true of most companies which require translation for their products, according to Rob Vandenberg (@robvandenberg) .
Translators, from what I could glean, are more interested in how software helps them establish workflows and manage their translation tasks.
But it’s undeniable that a TM could make a huge difference to what translators could do.
Lingotek has a really very neat interface that let’s translators organize workflow and manage translations in a very user-friendly way.
The software chunks up a text for translation, but offers context so you always know where a sentence fits. It also shows TM matches, automated translation results and terminology hits all in one place.
Translators or project managers can then go and change the text segments that will resolve to the TM according to what makes most sense for them. Similarly, you can filter TM matches to level of accuracy you choose.
This all points to the crucial need for flexibility. But it also suggests that – as far as corporate texts with a high degree of textual repetition go – TMs and terminology managers can really help translators improve productivity by offering up segments and terms that have been translated before.
WeLocalize’s Global Sight platform has similar functionality, but perhaps lacks the crystal clear interface. What the platform lacks in UI design it makes up for in flexibility, as the tool really offers a powerful way to manage translation projects.
The software has three project roles – project manager, translator and reviewer – whose duties and permissions can be configured for each project.
Globalsight also calculates cost for translations, allowing you to customize each activity based on a per hour or per word rate in whichever currency you choose. Shame there’s no Paypal or Kiva integration in there though.
The Rosetta Foundation, an off-shoot of the Irish government-supported Centre for Next Generation Localization which brings together 100 researchers from Limerick University and three Dublin research centers, seeks to provide Translators Without Borders with the technology it needs to make access to information more equal across languages.
As this was the first ever AGIS, there were impassioned speeches in favor of better access to information.
But, perhaps unexpectedly, the chief organizer of the event, Reinhard Schaler, made a slightly different point. He urged the localization and software industries to think that, whether they like it or not, the market is moving south and east, and fast.
He proposed that the tools and knowledge sharing platforms for this next generation of localization would have to take account of many languages which are currently completely off the radar.
That’s a point that Asia Online – represented here by the inimitable Kirti Vashee – has not only run with, but built an entire business model around.
If the challenges facing Meedan seem vast in our efforts to bring social translation and emerging machine translation technologies to the task of better dialogue between Arabic and English, spare a thought for Asia Online.
This is a company which seeks to not only translate huge portions of Wikipedia into East Asian languages including Thai and Indonesian Bahasa (which currently languish on the web), but seeks to create winning Statistical Machine Translation engines for these languages, in so doing kick starting a renaissance for content generation in East Asia.
The ultimate business proposition is to sell services along the way, with at its core the idea that Asia Online could pip Google, Yahoo and a cohort of local search engines to creating the best search for this fast emerging market.
That explains why Asia Online blocks Google and others from crawling its translation memories and corpora.
But the figures for achieving this seem daunting. Vashee imagines it would take 500 million sentences to produce accurate, near human statistical machine translation.
Before that happens, a complex process of training its MT engines has to take place, involving different tiers of community involvement.
Still, you have to respect the ambition of this startup – which seems to wed a vision for changing the world with a business model to match.
Now there’s something for us all to chew on.