Translation is not an easy game to play. One wrong word, or the right word in the wrong place, can flip your meaning upside-down. The opportunities for error stack one atop another. Reliable automatic translation using Google’s techniques is not just a dream. The way language works, it is a mathematical impossibility.
Universal translation is a dream, like the flying car. Yet several myths lead people to believe Google has achieved it already. Artificial Intelligence. Neural Networks. Zero-Shot Translation. With these buzzwords and more, we think that Google has solved language. This chapter looks at the myths, and looks at what really happens.
Does Google Translate translate? You might put your faith in it. You might laugh at it. I studied it, for all its languages. This is the first comprehensive examination ever conducted of the world’s most-used translation tool.
In some situations, GT is an excellent tool. In some, it is helpful as part of a broader translation strategy. In many, though, it should be avoided or cannot be used. Based on this study, here is a guide on how to use Google Translate, and when to run, run, run, run away.
Facts matter. Google Translate processes 100 billion words a day. In all languages some of the time, and in most languages most of the time, analysis deep inside the service concludes: the operating mode is to make stuff up.
I conducted a scientific evaluation of translations produced by Google, for all 102 languages vs. English in their roster. Such a study, using native speakers for every language, has never been done before. Evaluators everywhere from Samoa to Uzbekistan examined Google translations in their language. Here is what they reported.
The missing link in using machines to produce great translations is, ironically, people. Language resides in our heads, so we must transmit that data to our computers before the computers can learn how to use it. This has not been done systematically, even for the few languages that Google and others have invested in heavily. Moreover, we need new ways to guide our machines through our words and towards our meaning. There is a path that could work for every language.