After learning of the Localization World keynote address topic, I decided to attend the conference in an attempt to identify why our industry leaders chose a theme this year that flirts with commoditization and inferior quality, when everyone else avoids them.
Coined by Jeff Howe of Wired magazine, Wekipedia defines it as follows: "Crowdsourcing is a neologism for a business model in which a company or institution takes a job traditionally performed by a designated agent (usually an employee) and outsources it to an undefined, generally large group of people in the form of an open call over the Internet."
Jeff gave examples on how iPhotoStock and Wikipedia were able to reduce costs by creating a virtual workforce made up of volunteers and industry enthusiasts that perform small chunks of work at their leisure virtually for free, or 10-100 times cost reduction.
The difference between outsourcing and crowdsourcing, is that instead of outsourcing work to few selectively chosen and paid professionals, you divide up the work into many small chunks and tap into a high volume of amateurs (first crowd), debutants (second crowd) or undiscovered talents (third crowd), that will do your work for you at a much reduced cost.
With the advent of free web technologies, making work available online is not the real challenge. The challenge is in investing time and resources building outsourcing communities by weeding through vast crowds, identifying the hidden talents and intrinsically motivating them to do work for you, and [with a drum-roll please] virtually for free.
Hmm, would you please tell me why would professionals in a high demand market such as translation, do the work virtually for free? Jeff's answer was that the trick is to identify the third crowd that likes to do it for self-satisfaction or recognition.
But translation is intensive and laborious work and unless one gets something in return, not many get enough joy from it to do it for free!
If you live in a fantasy world where localization is not a must-have, time is on your side, quality is not necessarily a requirement, company brand and image are a non-issue, and you have resources to invest in wikis and source management, then Crowdsourcing may be your answer.
For the rest of us that live in the real world, where quality and timely localization is a live or die endeavor, there is no substitute to hiring professionals and paying them their earned dues. Anyone that tells you otherwise is fooling you, regardless of your age!
Saturday, October 27, 2007
Is crowdsourcing localization an option?
Subscribe to:
Post Comments (Atom)
Posts
2 comments:
Nabil,
I understand your thrust. It seems to me that the newly recognized potential in crowdsourcing is that the crowd can provide an answer to quality as well as quantity, and potentially to speed as well. In particular, while only a few people may contribute, a lot of people would be in a position to rate them. This woulc happen either by explicitly voting or implicitly by just using the results. The higher quality ones will potentially (certainly not guaranteed in any old community) drift to the top. As you said, the challenge is in creating, even farming, the right community, which contains both experts and caring amateurs, who's total effort results in something unexpectedly good. This doesn't mean it's the same as what a single professional would have created, but it means the users are satisfied.
Speed would be a function of a couple things: 1. the ratio of the participating community to work -- ie if there's lots of work for few people, it won't be fast; 2. the means of attaching value that's used -- eg the mturk.com approach of letting the work supplier define the price to attract workers, vs community-based prioritization -- eg certain projects like translating Al Gore's latest global warming speech might be considered "important" by lots of people, which drives their volunteer efforts (and pulls them away from other projects).
Now for a "customer" trying to get something arbitrary/obscure/unpopular translated, this would still be a very different way of doing business than hiring a professional and getting assurances from that one person. This suggests that it will be as dependent on avant garde customers figuring out appropriate ways to make use of such a service as it develops, as much as the service itself maturing and being able to set clear expectations. But in the long-term both quality and speed could potentially be solved.
For localization (of software in particular) of course there's both domain expertise and technical complexity, so you'd need a somewhat different community participating.
Thoughts ?
While you are right about general localization activities being difficult to do via crowdsourcing, there are cases where it works even for that. Companies interested in clearly defined high quality on a specific schedule will generally prefer to go to a competent translation service provider. There are very few people interested in helping to translate a printer manual, but they will help to translate things they care about e.g. travel and hotel reviews, opinions on global warming, green technology, religion etc...
Some examples where crowdsourcing is working even for localization:
Google
Facebook
Second Life
Microsoft
Facebook’s mantra of efficiency and effectiveness has been crucial in its ability to grow outside the US and other English speaking countries. Chandlee explains that global growth has been pursued not through infrastructure expansion in other markets, through translation. One of Zuckerberg’s innovative stances on the web is that users are not defined by their nationality, but rather by the language they speak. “Rather than launching Facebook in France, we launched it in French.” This in itself is a novel way of thinking… and the method they used for local launches is quite astounding. Facebook set up a translation application within the site which 3000 Spanish speakers signed up to. The app requested each user to translate an English line into Spanish. Then all the users would vote for the best translation of that line. The most popular line became the Spanish version. This process was repeated for a number of lines till the site was fully translated. Wiki-Spanish.
By crowdsourcing its own user base, it took 3 weeks to translate Facebook into Spanish.
German took 6 days.
French took 6 hours.
If you have a motivated user community it is possible to do this. Even Microsoft who spends $300M+ on LSPs each year is using "managed crowdsourcing" to clean up the quality of the MT output in the knowledge base. There is no way that this could be done without automation. Again they have a large and motivated user base. What they need to do to capitalize on this is manage it to ensure quality and standardization. They are doing this through their MVP partners.
Crowdsourcing is about direct contact with the global customer.
I think it is wiser to understand the phenomenon than simply condemn it, since one can envision a few scenarios where it does not make sense.
I think there is a definite legitimacy to crowdsourcing (which is not always free) as this is the only possible way to convert the large masses of knowledge in the world to languages where people are information impoverished. In 2006, the amount of digital information created, captured, and replicated was 1,288 x 1018 bits. In computer parlance, that's 161 exabytes or 161 billion gigabytes … This is about 3 million times the information in all the books ever written. It is not acceptable to wait for translators to get around to translating what people are willing to pay for. Information should not be the right of those who speak a G-7 language.
MT is also necessary. However, being in a world where Google controls this is less appetizing. Man machine collaborations are necessary to raise the quality of MT, as this is what evolution demands. Hopefully others in the industry and translators together develop win-win scenarios to preempt the Google effort to dominate this area.
Just as cars and planes get you places faster, automated translation technology will make it possible to get large amounts of content translated faster. Technology is not a replacement for humans, it can however, make them much more productive and some will learn to use technology and social networking phenomenon like crowdsourcing to things that were considered impossible before.
I think this is actually an opportunity for LSPs who learn to use technology to enable massive amounts of dynamic content to be translated. Some LSPs will be at the leading edge in terms of learning how to properly use SMT and massive online collaboration to get things done faster and cheaper.
I predict we will see many more examples in 2009 where automated translation technology together with massive online collaboration will change the way large scale translation projects are done.
Post a Comment