‘How do we know what to translate?’ Notes from OTT09, Amsterdam
Ivan of Global Voices speaks with Ed of the Meedan project at OTT09 in Amsterdam. Picture by oso.
If there’s one thing the participants of last week’s Open Translation Tools conference in Amsterdam agree on, it is surely that a polyglot web means a richer, more diverse, more equitable web where access to knowledge and resources is more equally spread.
To quote Ethan Zuckerman on the power of the polyglot web:
For the the Internet to fulfill its most ambitious promises, we need to recognize translation as one of the core challenges to an open, shared and collectively governed internet. Many of us share a vision of the Internet as a place where the good ideas of any person in any country can influence thought and opinion around the world. This vision can only be realized if we accept the challenge of a polyglot internet and build tools and systems to bridge and translate between the hundreds of languages represented online.
But unpacking the broad idealism of the translation agenda at the conference poses two slightly distinct motivations.
One is about absolute need: making sure that web users are not completely excluded because of their language, or as Zuckerman puts it, ‘making tools and knowledge accessible to a global audience’.
Dwayne Bailey’s Translate.org.za, for example, seeks to translate open source software that many English speakers take for granted into the 11 official languages of South Africa.
Without localization of core knowledge and tools to navigate and contribute to the web, users from minority languages will have little option but to learn a language with global reach, like English or Chinese.
But beyond the localization of core knowledge and tools is a broader agenda to enable what Zuckerman calls cultural ‘bridging’.
This is where Meedan sees its role in the evolving polyglot web. It is about sharing perspectives, experiences, a broader sense of ‘knowledge’.
Anas Tawileh demonstrates Meedan.net to translation delegates at OTT09. From itzpapalotl's photostream
As Ethan Zuckerman notes: ‘If you want to know what people around the world are thinking and feeling, you need help from a translator.’
But this kind of translation is not only challenging in terms of scale and scope, it is challenging simply by virtue of the apparent paradox it presents.
As Ed Bice asked a discussion group on the subject at OTT09, how do we aggregate audience demand when readers don’t understand content in order to choose?
There surely has to be some form of pre-selection of content going on.
The dominant model is for publishers to determine what should be translated. For commercial outfits, this is way of reaching new markets.
The Iraqi news agency Aswat al Iraq has both an English and an Arabic language wire service, for example. A selection of the stories deemed appropriate for an English language audience are translated into English by a translation agency in Egypt. Many Middle East news outfits do something similar – Asharq Alawsat, Al Hayat, Al Masry Al Youm, and Al Arabiya to name but a few.
For nonprofits translation can be a necessary task to achieve broader mission.
Search for Common Ground’s excellent News Service – which combines reports, comment and analysis pieces in Arabic and English – is a highly successful example.
Elsewhere, translators say what they think should be translated.
The Chinese social translation hub, Yeeyan.com, has a mechanism to enable users with language skills to recommend English language articles for translation into Chinese languages.
But even here, there may be a mismatch between what a translator is interested in and what a broader reading public is interested in. Yeeyan’s most successful project has actually been to translate The Guardian rather than through recommendations systems.
Wall poster of Meedan's hybrid translation workflow, presented by Anas Tawileh at OTT09. Picture from itzpapalotl's photostream.
If we could see statistics for pages translated by Machine Translation services, like Google Translate we could begin to gauge interest in articles by users of other languages.
Alternatively, MT services could be used as a gist reading tool to enable a broad set of users to collaboratively recommend articles for human translation based on the rough semantic outline provided by MT.
This sounds plausible enough, but Jiamin Zhao of Yeeyan.com suggests this approach is not as compelling to users as a recommendations system because readers are not willing to wade through MT.
Meedan’s approach is to enable translators, content producers, and users to recommend and share links around key news topics – which we call ‘events’.
The motivation for Meedan’s approach was that users from different language communities – who otherwise have little opportunity to interact – may share significant experiences through world news events.
So when Barack Obama gave one of the most challenging speeches ever given by a US President at Cairo University, we knew that American English speakers would have a reason to hear from Middle Eastern Arabic speakers and vice-versa.
The result was a new narrative of cross-language cross-cultural engagement, where users from across the two language groups shared responses and articles in translation.
This is one of the central observations of Meedan – that communities of interest and communities of language coalesce around time relevant events.
Common events present us with opportunities for engagement. Despite Iran’s insistence that the post election protests were purely an internal matter, the unprecedented explosion of citizen reporting on Twitter suggested this was not so.
The State Department’s intervention to delay Twitter’s downtime and Google’s sudden move to release its Persian MT service well ahead of schedule only go to reinforce that this was a citizen-driven media event in which more than ever before web users were encountering – and even seeking out – content in other languages.
There do exist systems for translating specific content – such as Twitrans for tweets and Lingtastic for SMS.
But these outfits lack community translation components that allow for a view of collective interest in content for translation, and social translation tools for volunteers to help out with the actual translation work.
At Open Translation Tools, we discussed the idea of a Firefox plugin that could allow you to browse the web in translation and request human edits of specific articles.
In the absence of this tool right now, maybe one way to start the ball rolling would be to use existing tools for sharing interest.
This is after all one of the driving forces behind the social web: here’s a link that I’m interested in, and you should read it too.
Twitter, Delicious, Digg, Diigo, Facebook et al are all about this.
So maybe we could generalize tags across these platforms that allow users to say, here’s a link that I’m interested in, and I’d love someone to translate it.
So, en2ar? means I’m requesting a translation of this link from English to Arabic.
And, #en2ar means I’ve translated this link from English to Arabic, check it out.
It would be fantastic if we could generalize these tags across the user community, call on the platforms themselves to standardize them in their FAQs, and get people using them.
You could even imagine a platform for sharing the resulting feeds, with the potential to cluster common links, offer bounties, enable micropayments to translators and connect translators and publishers.
All these ideas may live in the future – but the folks at Open Translation Tools 2009 were among those leading the way there. Great projects like DotSub, Worldwide Lexicon, Global Voices and Traduxio show that for sure.
Still, if you want to come and help share and translate links about the Middle East, you’re welcome at Meedan any time.
