Re^4: X-Prize: Natural Language Processing

Maybe the goal for the NLP X-Prize should be a little more ... constrained. I was thinking about constraints this evening. What about the following change?

A successful NLP X-Prize program should be able to translate any paragraph that is:

self-contained with respect to context
restricted to a given wordlist (maybe 1000-2500 of the most common words)

and translate that paragraph from any one of N languages to any other of those languages. The languages would be set in the final specification, but would include the following:

English
French
Chinese (written)
Japanese (Kanji, probably)
Hindi
German
Russian
Arabic
Navajo (or some other AmerIndian language)

The program would perform the translation under 1 second per word of input or output, whichever is greater.

The wordlist would be chosen to be dialect-agnostic, as much as is possible. Most of these words would be the words most people learn in grammar school. I'm talking about words like "in", "is", "have", "run", "cat", "dog", etc.

The X-Prize wouldn't go to the program that can translate physics texts, just like the Ansari prize didn't go to the ship that would actually ferry passengers. It went to the ship that demonstrated the feasibility of technologies. Now that SpaceShipOne has succeeded, new ships will be built to actually make it a commercial venture. I would expect the same to happen for any other X-Prize.

Being right, does not endow the right to be rude; politeness costs nothing.
Being unknowing, is not the same as being stupid.
Expressing a contrary opinion, whether to the individual or the group, is more often a sign of deeper thought than of cantankerous belligerence.
Do not mistake your goals as the only goals; your opinion as the only opinion; your confidence as correctness. Saying you know better is not the same as explaining you know better.

Comment on Re^4: X-Prize: Natural Language Processing