MadSci Network: Computer Science
Query:

Re: Is it possible to code a Universal Translater?

Date: Wed Sep 29 13:13:17 1999
Posted By: Ryan Scherle, Grad student, Computer Science & Cognitive Science, Indiana University
Area of science: Computer Science
ID: 937924316.Cs
Message:

Hello Edward,

This sort of concept is "almost" possible now. We already have portable language dictionaries, so you can look up a word that you don't understand, but this requires you to know something about the language you're using. You could buy a laptop computer and equip it with software to do translation, but the current state of the art wouldn't give you very good results.

There are two major pieces that this type of translator needs to work properly. First, it would need to do speech recognition (listening to a voice and deciding what words were said). Second, it would need to do the actual translation to convert the words from one language to another.

Speech recognition is reasonably reliable now. You can now buy a speech recognizer at any software store. The best English recognizers can generate 90% accuracy for a voice they are familiar with. Recognizers for other European languages should be comparable, but some other languages (eg. Chinese, Swahili) are very different from English, and recognizers may have more trouble with those languages. Of course, if you're walking around carrying a translator, it won't be familiar with the voices that you come in contact with, so the accuracy will be quite a bit lower.

Machine translation is an equally difficult problem. You can see the a good representation of the current state of the art at Alta Vista's site. Translation is more than just looking words up in a dictionary. In order to do a good translation, the computer must be able to have some understanding of the sentence. If the computer doesn't understand the sentence, it can't always pick the correct translation for a given word. Machine understanding has been a focus of artificial intelligence research for decades, and we still have a lot of work to do.

When you combine the errors produced by the speech recognizer with the errors of translation, the results may not be recognizable.

In the future, both of these fields will improve, but it will take time. In the next decade, we may see translators that work for one or two languages, and do a reasonable job of getting the main point across. However, they won't rival human translators for the next 50 years or longer.

For some more information on speech recognition and translation, look here.


Current Queue | Current Queue for Computer Science | Computer Science archives

Try the links in the MadSci Library for more information on Computer Science.



MadSci Home | Information | Search | Random Knowledge Generator | MadSci Archives | Mad Library | MAD Labs | MAD FAQs | Ask a ? | Join Us! | Help Support MadSci


MadSci Network, webadmin@www.madsci.org
© 1995-1999. All rights reserved.