The Universal Translator Assistant Project

using the technology of today to bring the theories of yesterday to the languages of tomorrow


The Atlantic Magazine Has Published some *EXCELLENT* articles about machine translation - read them here:


Theory Behind the UTA

Not much. Call it a science fair project gone awry.

Used as a lookup tool, it lets you layout vocabulary words for an aid in translating a text. As a translator, it demonstrates the weakness of machine translation of natural language texts.

A while back, I acquired a machine-readable copy of the WEB Bible translation (http://www.ebible.org/bible/WEB/webfaq.htm ) and was doing a number of text processing things with it. Here are some programs to use  if you'd like to use the WEB Bible (or King James, now):

are simple programs I wrote to look at it, on PCs - or any machine with a C compiler - the two "webship" archives contain C source.)

Along the way I realized it was easy to create a list of all the unique words in a passage. From there I wondered "what would happen if I took such a list, say from Isaiah 1, and (free of context) mapped each word (that I could) to a Klingon word. With such a mapping, I could quickly lay out the words in a verse, and reshape them into a grammatical Klingon translation.

Here is a simple diagram of the process:

There are three steps to programming UTA
1. Choose a source language text to use.

2 Reduce it to a list of all unique words.

3. Translate each word into the target language to have a word-for-word mapping

Then it can be used to (badly) translate between the two languages, after a fashion.

[Image]
Compare this to the far more complex Universal Translator, as described in the Star Trek The Next Generation Technical manual: [Image]

What was particularly interesting was the result of using this to 'translate' a phrase into Klingon, and then back to English. While it (of course) renders an atrocious meaningless Klingon phrase. I found that the back-translation resembled the broken English of a person learning English as a second language. While of limited practical use, it demonstrated to me that I was indeed modeling one form of language mapping that occurs among humans.


* How can I use UTA to translate English into Klingon (or other languages)

You can't. Don't try. What UTA produces is coded English - treating the target language as a code of English. This IS a way that natural languages get used (or abused) by non-native speakers of a language - but it is decidely NOT a translation into the target language.

For real information regarding the Klingon languages refer to the excellent materials from Marc Okrand (The Klingon Dictionary, The Klingon Way, and Klingon for the Galactic Traveler) and visit the web sites of:

Let me note here that the vocabulary I provide is a highly dubious one - I really forced some words in order to come up with a Klingon  word to match English words.


What is wrong with this?

The UTA program does not - CANNOT translate. All it does is map words from one language to another.  Early in the history of MT (machine translations) this idea was easily disproved.  Here is an experiment - use UTA to "translate"

I do not like you

into Klingon.  You will get:

jIH ta' ghobe' rur SoH  

That seems great, and it almost translates back exactly.  Translate it back and you get "I do neither like you".

No one  of the translation words is wrong:

But it doesn't fit together at all as grammatical Klingon.  Besides the meaning for "like" UTA has is "resemble", not the sense of affection or  friendship.

The best this could mean in tlhIngan Hol is

"He accomplishes I. You resemble no."

More likely it would be seen as what it is: NONSENSE. True Klingon for "I do not like you" is far simpler:

qapar : qa- (I-you) par (dislike)

Grammar *really* does matter. UTA  can be a good program, for amusement, or automated word lookup. NOT translations.


Back to the Universal Translator Assistant