Lost in Translation

Linguistic translation is far from perfect, especially when dealing with translations between eastern and western languages. I recently performed a simple experiment on the Altavista translation software, Babelfish, to show the inferiority of computers in interpreting natural language and the inevitable event of “losing integrity” when translating from one language to another. If you didn’t know, a “Babelfish” is a fictional piece of biological technology from Douglas Adams “A Hitchhikers Guide to the Galaxy” which crawls into the recipients ear and into their brain, allowing them to understand all languages of the universe instantly.

Natural language is full of so many oddities, accents, tonal variations, homonyms, intuitive meanings, etc — it appears as if a mathematical classification of such a system is impossible without integrating the concept of “the speaker” into the equation. By this I mean there are important clues to the meaning and intent of the language that cannot be expressed with the words themselves, but must be conveyed by the speaker physically — with tone, accents, and body language.

The Experiment: This is pretty simple, but can lead to some very interesting results that expose the ineffectiveness of online translation tools. The two languages used for this example are English and Japanese. Take a simple English sentence — translate it to Japanese — translate the Japanese back to English. Take the new English result, translate it back to Japanese. Wash, lather, repeat.

Here are the results for the seed string:

This sentence will not translate very well into Japanese. It will not translate into any eastern language for that matter. The languages are just fundamentally different.

Here are three subsequent interations of the translation:

It does not translate this sentence very well in Japanese. At this point in time it does not translate that in language of all east. Language differs fairness basically

That does not translate this sentence very well with Japanese. At this point in time that does not translate that of language of all east in time. As for language fairness it differs to the basis

That does not translate Japanese and this sentence very well. At the time when it does not translate that of language of all east of time at this point in time. Language fairness it differs to the foundation to in regard

Notice the redundancies that build up on the text after each translation? Where did the concept of fairness come into play? The algorithm used for this system is primitive at best — after 100 iterations the original meaning would be lost entirely. A perfect translation is not possible — the axioms that underly the words are fundamentally different between the two cultures. Converting between the two is like squeezing a square through a round hole — pieces of the original idea will be lost.

Here’s the important point: Historically and religiously important documents such as The Bible, the Tao Te Ching, the Koran, etc — all of them are translated into English, sometimes through several different languages before finally arriving on a final English version that ends up in your local Barnes and Noble. How can we trust these documents, or base entire religions on them for that matter, when the original meaning has been shaven off, piece by piece, leaving only a superficial representation of their true intent. This is akin to playing telephone with the words of Mohammad, Jesus, and [insert influencial thinker here] through thousands of generations of people. Each time the word is passed on it changes.

Perhaps one day a true “babelfish” will exist. Perhaps we will be able to reach a new form of communication that is beyond simple words and reaches into something truly universal and intrinsic in the human being.

Wow, that was a lot of rambling and pretensious word-play for a Saturday night. I need a beer.