AlfaBetic is a translation service that offers publishers a free way to have their content translated to ten of the web’s most popular languages, which it says will expose them 842 million potential new readers.
The service first runs text through a computer, which is similar to the web-based translation systems offered by Google Translate and a number of other sites. After generating this rough translation, AlfaBetic then presents the translated text to paid human translators, who ensure that everything makes sense. Alphabetic says that over time it will be able to reduce its costs by making the original automatic text more accurate using “statistical machine translation”.
The service’s most impressive feature is its ability to keep comments consistent across every translated site, as well as the publisher’s original blog – I could leave a comment on TechCrunch Russia and have it appear on TechCrunch.com, translated to English. These comments are still translated by machine and read by a human, but the company says that the process only takes a few minutes.
AlfaBetic will control advertising appearing on each publisher’s translated pages, and says that it will try to maximize revenues by using local ad merchants – something that might work well in some European countries, but markets in countries like China and Russia are far less appealing. The company will take a share of this ad revenue, but the service will be free. It sounds like distribution models for content will depend on the publisher, but at the moment AlfaBetic says that it will act as a syndication service, posting the material to a number of different outlets (as well as its own server).
Discussion
Om Malik – Whose translation engine do you use?
Alfabetic – We developed our own translation engine, have been working on it for a couple of years, special emphasis on different domains of language.. Emphasis on spoken, politics, etc. We can focus on each domian.
Om Malik – What’s your accuracy?
Alfabetic – depends… we believe 90%?? Politics for example 80-90%
Om Malik – How many ads can you really serve against this content? It seems like bloggers don’t have control over how many ads go in there.
Alfabetic – We’re offering the publishers. a complete turnkey solution. That means that we will do our best to create an ad network that will maximize the revenue, and the best way to do that when you translate, is to partner with local ad aggregators, get them from different local publishers, ad providers.
Tim O’Reilly – So you guys have actual experience, you’ve done this for TechCrunch. How did this work? Ad revenue?
Alfabetic – What we did was a pilot, it was not open to public. We believe that if we translate into 10 most popular languages, we don’t know if readership per language will be as large as in English…
Tim O’Reilly- what’s the infrastructure? How is this hosted? Does Michael (Arrington) have to get a domain name?
Alfabetic – We were thinking of a few models. Different blogs, different publishers fit different models. If we could partner with popular portals in each language.
Tim O’Reilly – So you become an ap to feed into other stuff.. syndication.
Alfabetic – Yep
Tim O’Reilly – What about world wide lexicon, a similar project – they dont do the monetization part.
Alfabetic – We know Google has big stake in this tech. We decided to concentrate on what we can do best. Train computers to translate particular domains.
Tim O’Reilly – Are you crowdsourcing the translation?
Alfabetic – we have people that we pay around the world to do the proofing.
Tim O’Reilly – Scaling problem, you have to cherry pick who you are going to work with. Who pays your?
Alfabetic – We get paid by ad monetization.
Alfabetic – We are about a tenth of what it costs to us e a translation agency. For us to translate techcrunch for us.. per month is 1k in all languages.
Josh Kopelman – We heard MySpace say 9 countries represent 90% of their advertising.. aren’t rates lower elsewhere?
Alfabetic – Europe would be very interested, like Germany. China and Russia and other regiions in world.. the numbers cover up for the economy. We’re looking towards the future
Evan Williams – Would Tim and Om look at this?
Om Malik – Probably not. If we were going to go international, we’d have the same model as what we have., we’d want dedicated stories. Translation works for very large media companies.
Tim O’Reilly – Probably not. One of our experiences, book publishes is a relatively small part of markte, yuo have have to have something local. You have biz dev issues.
Video
AlfaBetic presented during Session 8 of the TechCrunch50 conference. You can watch its stage presentation below.