Posts

Google and IBM Tackle Ambitious Translation Projects

You may not realise it, but the internet sites you see when you surf the web are limited by the language you are surfing in. So, there’s always a part of the internet that you are shut out of.

Possibly, according to this article on CNN.com, not for long, both IBM and Google are currently working on ambitious translation projects. If successful, they will be able to create accurate, instantaneous translations of web content.

IBM’s project, called n.Fluent, automatically translates web pages. Users can enter in a URL to get the page translation, and site owners can install a translation web app onto their site that allows users to choose their language from a drop-down menu to have the page translated.

Of course, machine translations are still imperfect-a problem IBM is trying to rectify by crowd-sourcing the work, tapping its multilingual workforce to improve the translation software’s capabilities. The approach has helped IBM rapidly improve the quality of their translations, but David Lubensky, an IBM “real-time translation” specialist, told CNN that the company still faces challenges with this approach:

“There are two challenges. Firstly, getting a sustainable, enthusiastic community can be difficult. The goal is to have an ongoing interest, to make it part of the fabric.  The second issue is quality assurance of content; how useful is the feedback, how many mistakes do people make and how much impact will they have?”

Google’s project is similar, but naturally, the titans of search are also interested in translating search terms.  Google’s vice president Marissa Mayer explained the goal of her company’s translation tool to the Daily Telegraph:

“Imagine what it would be like if there was a tool built into the search engine which translated my search query into every language and then searched the entire world’s web sites.  And then invoked the translation software a second and third time — to not only then present the results in your native language, but then translated those sites in full when you clicked through.”

Although IBM’s project is aimed more at businesses and Google’s appears to be aimed more at consumers, both would go a long way toward opening up the entire internet to everyone. However, it’s unlikely that they will replace the services of an experienced translation company-if you are translating something important, you want a human being who understands the nuances of both languages to do it.

Translating the US State of the Union

Every year in late January, the President of the United States gives a speech to both houses of Congress, reporting on the “state of the union” and what lies ahead in the coming year. Of course, it’s not just the United States Congress listening to the speech. Interested people from across the globe listen, too. This year, to make the speech more accessible, news organizations and others turned to crowdsourced translation.

First, translation startup Babelverse announced that they would use their real-time translation service to livestream President Obama’s speech into as many different languages as possible. They hope to eventually offer livestreams like this in all 6,976 languages. While that goal proved to be a tad ambitious, according to their blog they were able to stream the speech into at least 7 different languages (Spanish, Hindi, French, Mandarin Chinese, Greek, Bahasa Indonesian and Portuguese), with an average of 2 interpreters per language.

Meanwhile, PBS, America’s public broadcasting service, brought back last year’s popular partnership with Universal Subtitles to offer crowdsourced subtitles for the video of Obama’s speech. Anybody capable of interpreting can volunteer with Universal Subtitles, and volunteer translators review the translations to catch errors.

A few hours after the speech was over, GigaOm reported that it had already been transcribed into 7 languages, and currently it is available in 29.

PBS NewsHour correspondent Hari Sreenivasan told Poynter.org that the project “gives viewers an opportunity to be part of spreading content to more people, and gives public media organizations a way to engage with their communities in a deep and ongoing way.”

Of course, even with a staff of volunteers, accuracy is always an issue. For now, at least, PBS is relying on its viewers’ better angels to keep its translations on target. As Sreenivasan put it:

“We have every intention to be as editorially accurate as possible. I don’t fundamentally believe that there are people out there who want to malign us by offering incorrect translations, but who knows? I think our intentions are noble, and I think the people who end up as volunteers for these kinds of things are generally more philanthropic and more volunteer-driven than the average viewer.”

The bottom line? Crowdsourced translation has a lot of potential for certain situations, but when getting it right the first time is what matters most, a reputable translation company is still the best way to go.

Image Source: AttributionNo Derivative Works Some rights reserved by KJGarbutt

Twitter Translation Center Adds More Languages

Twitter has been using a crowdsourced translation model to publish the site in languages from around the world. As of this week, the total has risen to 30 with the addition of Ukrainian and Catalan. Ukrainian is spoken by 37.5 million Ukrainians and an additional 3.5 to 7.5 people outside of the country.

Catalan is spoken by about 6.7 million people in and around Catalonia, Spain. It is a Romance language related to Spanish and Italian.

In a blog post, Twitter noted that its users have been clamoring to translate the popular website into their own languages:

“The demand has been so high that we built a console – Twitter’s Translation Center, where users can help suggest translations for the site. With each official Twitter language launch, we saw more and more demand from users to help us translate Twitter into their language.”

The success of Twitter’s crowdsourced translation efforts holds a lesson for businesses of all types: people relate better to products that are presented to them in the language they prefer to speak. But crowdsourcing translations is certainly not for everyone. Translations can take time to perfect using this method, so you need a dedicated community that will be willing to help with quality control and be patient with translation errors.

Here’s how the process works with Twitter:

If you see a translation that doesn’t feel right, the best way to fix it is to log into the Translation Center and vote for the best choice. The phrase score will adjust over time, and the right translation will find its way into Twitter.com. If you see inconsistencies in the content, you can get involved in the discussions with your community at the language forums. We’re continually improving on the system, so please send us your feedback and report any issues you come across.

This process works well for Twitter and Facebook because they have so many passionate users. For most businesses, a knowledgeable translation company is still going to be your best bet when it comes to moving in to new markets.