A Useful Translation Aid

THE Author of the Bible, Jehovah God, desires that the good news of his Kingdom be declared “to every nation and tribe and tongue and people.” (Revelation 14:6) He wants his written Word to be readily available to all mankind. To that end, the Bible has been translated into more languages than any other book in the world. Thousands of translators have spared no time or effort to render God’s thoughts into another tongue.

But the Bible is not simply the object of translation. Time and again, it itself has been used as an aid in translating other texts. Many a translator has compared the renderings of Bible terms in different languages to come up with a good translation for certain words. The Bible’s qualities as a translation aid have now been put to use in computer translation as well.

It is really difficult for a computer to translate. Some experts have even felt that translation is beyond what a computer can do. Why? A language is not just a set of words. Each language has its own word combinations, rules, exceptions to these rules, idioms, and allusions. Efforts to teach a computer all of this have met with very little success. Most of the resulting computer translations have been barely understandable.

Now, however, computer scientists are exploring new ways. “Our approach uses statistical models to find the most likely translation,” says Franz Josef Och, a leading specialist in computer translation. Let us say that you want to translate Hindi into English. First, take some existing text that is available in both languages. Then feed it into a computer. The computer compares the texts. When, for example, the computer finds the same Hindi word in several places and each time in the corresponding phrase it finds the English word “house,” the computer concludes that the Hindi word must be the word for “house.” And chances are that nearby words are adjectives, such as “big,”  “small,” “old,” or “new.” Hence, the computer builds a list of corresponding terms and word combinations. After such “training,” which might take just a few days or weeks, the computer can apply what it has “learned” to new text. While the resulting translation might be poor as to grammar and style, it is usually readable enough to convey the meaning and the important details.

The quality of the translation depends to a large degree on the quantity and quality of the text that was initially fed into the computer. And here is where the Bible comes into its own. It has been carefully translated into scores of languages, is readily available, and contains a sizable amount of text. So the Bible was the researcher’s first choice when training the computer for a new language.