Is AI Already Influencing Translation? Technology and Artificial Intelligence (AI) in particular, is unashamedly breaking the language barrier,...
3 min read
09/01/2015
Several companies are working in machine translation and, in fact, some large companies are "BUYING" machine translation companies so that the technology becomes part of their offering a it becomes core to their business. Ebay did so with Apptek in 2014 and Facebook did it a year earlier, acquiring Mobile Technologies, the Pittsburgh-based developer behind the app Jibbigo. Previously, Facebook had been relying on Microsoft Bing for machine translation. Microsoft has gone a step forward and Microsoft Skype Machine Translator is finally here. If your business interests are based in more than one country, you probably want to have a certain level of ownership about your multilingual content and how it is published. You can translate yourself. Perhaps you are willing to pay for professional translations for your most relevant content, but leave not-so-relevant content to machine translation, which can happen on-the-fly. Furthermore, many of your transactions and interactions may be with foreign buyers and consumers... so perhaps it is not such a good idea to rely on "data sucking" companies. Many translators already use the services of generalist machine translation providers "with feedback" so that eventually. Fears of eavesdropping and telephone tapping, a major concern over the last few years are making more and more business edgy about how much of their data is being used by third-parties and for what (usually commercial) purposes. I should write more about how some users have had their AdWords campaigns stopped when they clashed against major accounts. Will Microsoft's Universal Skype Machine Translator be another way of "data sucking" for large software companies?
Learning foreign languages a thing of the past and "data sucking" the present?[/caption] At the end of the day, all digital companies publish data on the Internet, and most translate, having to offer content into several languages. Anybody publishing on the Internet makes his/her content available to public, worldwide scrutiny. Who doesn't (try to) use the Internet to operate worldwide nowadays, from buying software to clothes and read news and blogs from all over the world? Microsoft has had a deep interest in the language area for a long time, perhaps in the wake of Google Translate taking a lead in the area. Chris Wendt’s work at Microsoft is visionary and I personally proud both Pangeanic and Microsoft exchanged data and pushed the technology forward as founding members of TAUS. We have been also lucky to co-present with him several times at Machine Translation fora. Microsoft is creating something new with these two (existing) technologies. Both voice recognition and machine translation are not new technologies. They have existed in several shapes for some time, but they are now entering maturity stage. The problem was always to combine them together as each statistical system introduces a margin of error. It was not long ago that we wrote on Pangeanic's blog about Microsoft’s skype translator and how long it would take for it to really develop, assuming a much longer time since the first test into German sounded a bit stiff. The distance between German and English is much greater than between Spanish and English. My guess is that cracking machine translation will come first and Microsoft has picked the easy prey with English and Spanish. Google is not having such a good time with Indian languages and the Indian government is investing in its own machine translation technology. Adding voice means adding another layer of algorithms and statistics to the technology. Any word that is not recognized well will be rendered wrongly into a query sentence for machine translation. And if the starting sentence makes little sentence as it has not been understood from voice to writing properly, the translation can only get worse. Also, how Microsoft plans to make money out of this remains the question... Why would anyone speak to a friend whose language cannot understand? There may be some situations where the software will come handy to discern, just like Google Translate is useful in very concrete situations and for vocabulary and sentence query. I can imagine certain useful situations, like speaking to foreign mother-in-laws and relatives. But is this enough to justify the investment? Will people pay a subscription service for Skype translations? I don't think so. Therefore, the business case must be somewhere else. And considering how massive amounts of data (Big data) need to be translated, questions on eavesdropping, private/personal data and security that come to mind. http://www.newsweek.com/technology-makes-language-fluency-snap-292077 http://www.techtimes.com/articles/22266/20141216/skype-now-offers-real-time-translation-heres-how-skype-translator-preview-works.htm
[caption id="attachment_1486" align="alignright" width="261"]
Learning foreign languages a thing of the past and "data sucking" the present?[/caption] At the end of the day, all digital companies publish data on the Internet, and most translate, having to offer content into several languages. Anybody publishing on the Internet makes his/her content available to public, worldwide scrutiny. Who doesn't (try to) use the Internet to operate worldwide nowadays, from buying software to clothes and read news and blogs from all over the world? Microsoft has had a deep interest in the language area for a long time, perhaps in the wake of Google Translate taking a lead in the area. Chris Wendt’s work at Microsoft is visionary and I personally proud both Pangeanic and Microsoft exchanged data and pushed the technology forward as founding members of TAUS. We have been also lucky to co-present with him several times at Machine Translation fora. Microsoft is creating something new with these two (existing) technologies. Both voice recognition and machine translation are not new technologies. They have existed in several shapes for some time, but they are now entering maturity stage. The problem was always to combine them together as each statistical system introduces a margin of error. It was not long ago that we wrote on Pangeanic's blog about Microsoft’s skype translator and how long it would take for it to really develop, assuming a much longer time since the first test into German sounded a bit stiff. The distance between German and English is much greater than between Spanish and English. My guess is that cracking machine translation will come first and Microsoft has picked the easy prey with English and Spanish. Google is not having such a good time with Indian languages and the Indian government is investing in its own machine translation technology. Adding voice means adding another layer of algorithms and statistics to the technology. Any word that is not recognized well will be rendered wrongly into a query sentence for machine translation. And if the starting sentence makes little sentence as it has not been understood from voice to writing properly, the translation can only get worse. Also, how Microsoft plans to make money out of this remains the question... Why would anyone speak to a friend whose language cannot understand? There may be some situations where the software will come handy to discern, just like Google Translate is useful in very concrete situations and for vocabulary and sentence query. I can imagine certain useful situations, like speaking to foreign mother-in-laws and relatives. But is this enough to justify the investment? Will people pay a subscription service for Skype translations? I don't think so. Therefore, the business case must be somewhere else. And considering how massive amounts of data (Big data) need to be translated, questions on eavesdropping, private/personal data and security that come to mind. http://www.newsweek.com/technology-makes-language-fluency-snap-292077 http://www.techtimes.com/articles/22266/20141216/skype-now-offers-real-time-translation-heres-how-skype-translator-preview-works.htm
Next time you think languages, think Pangeanic Translation Services, Translation Technologies, Machine Translation