So there’s this project I’ve been working on for years which I’ve been (mostly) mum about.
No more. Now’s the time for talking — across borders, between languages, outside of our disconnected ecosystems of news-gathering.
Welcome to Meedan.
Meedan is a space for conversation and networking — the word ‘meedan’ (ميدان) means ‘town square’ or ‘gathering place’ in Arabic — where everything posted is mirrored between English and Arabic using a mix of human and machine translation.
The project is based on the simple (even self-evident) premise: it’s easy to distrust and misconstrue someone you can’t have a one-on-one conversation with. While the web is a place of massive social interaction, this interaction is almost universally bounded within language groups — a startling barrier to true understanding.
Meedan focuses on reducing this barrier by enabling English and Arabic speakers to
- share news and opinion from the English-language and Arabic-language web
- join cross-language conversations about technology, arts, business and politics
- widen their social network with people who speak a different language and who partake of very different cultures
- write, vet and edit translations in collaboration with users around the world
The project is led by the Meedan organization, a non-profit in San Francisco, with technical development and translation technologies from IBM. Here’s a video introducing Meedan.
So, how does it work?
Comments are instantly translated into Arabic or English using IBM’s machine translation. But because machine translation is not perfect (especially with a language as complex as Arabic) community translators are allowed to edit the translation.
This ability to improve the translations works like editing a Wikipedia article and, in my opinion, is the really novel use of social media on Meedan. (The plan is to allow translations to be rated such that, over time, the best translators emerge as part of a social network of trusted bilingual users.) As a final step, professional translators vet the community-submitted edits. Here’s a video demonstrating comment translation.
These hybrid machine-human translations are then fed back into the system which learns from the ever-growing, vetted corpus. The more people talk, the smarter the machine translation becomes.
Can the system be gamed? Sure. Will there still be misunderstanding, enmity, and deliberate mischievousness? Likely. You can’t change human nature. What Meedan does is provide tools for mitigating the less salutary effects of long-distance, networked conversation between peoples of different cultures.
That’s the hope, anyway. Meedan is in an open (though relatively quiet) beta phase right now. Come on in.