This week (Feb 2011), a computer named Watson, built by IBM over a 4-year period, won the 3-day Jeopardy! contest against 2 champions, Ken Jennings and Brad Rutter. With the word “Think” displayed prominently behind the contestants in many natural languages of the world, I couldn’t help but think what will be the artificial intelligence impact on translation.
What is Watson?
Named after IBM founder Thomas J. Watson, Watson is powered by a 2880-processor brain and runs algorithms based on Natural Language Processing (NLP). Briefly, NLP enables a computer to convert a human language into data that it can relate to. It accomplishes this by reducing letters, words, sounds or symbols into bits and bytes. It then matches them to a gigantic database of known information. In Watson’s case, this database is 15-terabyte strong. Developers are using basic NLP algorithms today in artificial intelligence based algorithms used in language translation programs like machine translation engines. You can also find them in spell-checking and grammar-checking utilities within many authoring tools.
We knew that computers can process data quickly and efficiently. Now however, the bar’s been significantly raised. IBM has proven in this week’s Jeopardy! contest, that it has the algorithms to make computers understand natural text. This is indeed a major breakthrough!
Effect on the Translation and Localization Industry
I often tell my clients that unless a translator understands your source text, there is little hope that he or she can translate or interpret it accurately into another language. This is true no matter how competent the translator is in his or her native language. GlobalVision relies heavily therefore on subject matter experts to perform translations. Ones that are competent in both the source and target languages.
IBM has proven that Watson can indeed understand a natural language, English, given many different subject fields. But can Watson accurately convey that meaning in another language? In other words, can a computer not only understand natural text? Can it also accurately compose the same meaning in any desired language?
Perhaps from what Watson demonstrated successfully this week, we can conclude that IBM now has the technology to ascertain if two sentences in two different languages mean the exact same thing. Since Watson can now understand an English sentence, it is not very hard to imagine IBM giving birth to another mammoth computer. One that they can call “Walid”. A sibling of Watson, that understands Arabic, or any other language for that matter. Then use the two computers’ algorithms and processing power to understand and deduce if the two sentences are equivalent in meaning.
This can be very useful to check if the quality of a translation, once available, is indeed a good one. But constructing a correctly structured sentence in another language to convey the same meaning as the source seems to be another piece of the puzzle. One that will remain to be solved. Perhaps it can be done, but requires yet another 3000-processor brain and many more years of further refined algorithms.
Artificial Intelligence Impact on Translation
Watson is a major breakthrough in the translation business, and technology continues its march toward automating translation tasks. But given Moore’s law, and assuming that it can be sustained for 20 more years, and that all the algorithms that are needed can be developed and optimized, it will still take at least a couple of decades before a computer can be built for use in a feasible commercial setting to replace the human translator.
Disappointed?! Don’t lose heart! Artificial Intelligence is making big strides. It is only a matter of time before language translation becomes a non-issue.
This white paper will guide you through the translation & localization process. It will show you how a reliable Translation Management System can alleviate many translation issues typically encountered in traditional translation and localization processes. Download it for free!