Mobile menu

KANT Translation System- Automatic Translation at Caterpillar
Thread poster: Maria Eugenia Farre

Maria Eugenia Farre  Identity Verified
Brazil
Local time: 20:16
English to Portuguese
+ ...
Mar 20, 2003


This is news for me.



http://www.lti.cs.cmu.edu/Research/Kant/PDF/lisanews.pdf



KANT is a knowledge-based translation system for multi-lingual information dissemination in pre-defined

technical domains. KANT has been deployed for French and Spanish translations at

Caterpillar, Inc. In this article, we give a brief technical summary, followed by a discussion of the

user’s perspective and experience with the system. We conclude with a discussion of ongoing work

and future plans for the KANT software.



The KANT system (Knowledge-based,

Accurate Natural-language Translation) is

a machine translation system for automatic

translation in technical domains, developed

at Carnegie Mellon University. KANT is

designed to allow the user to define the

terminology for specific domains, where

the meanings of terms are limited to those

relevant to a particular context (e.g., heavy

machinery, computer equipment, etc.); as a

result, KANT can achieve better

translations than general-purpose

translation systems when applied in

specific technical domains.





Does any one have more info about it or has actually seen this software at work?



Thanks in advance,



ME


Direct link Reply with quote
 

sylver  Identity Verified
Local time: 07:16
English to French
Some data... Mar 20, 2003

Quote:


On 2003-03-20 10:10, farre wrote:



This is news for me.



http://www.lti.cs.cmu.edu/Research/Kant/PDF/lisanews.pdf



KANT is a knowledge-based translation system for multi-lingual information dissemination in pre-defined

technical domains. KANT has been deployed for French and Spanish translations at

Caterpillar, Inc. In this article, we give a brief technical summary, followed by a discussion of the

user’s perspective and experience with the system. We conclude with a discussion of ongoing work

and future plans for the KANT software.



The KANT system (Knowledge-based,

Accurate Natural-language Translation) is

a machine translation system for automatic

translation in technical domains, developed

at Carnegie Mellon University. KANT is

designed to allow the user to define the

terminology for specific domains, where

the meanings of terms are limited to those

relevant to a particular context (e.g., heavy

machinery, computer equipment, etc.); as a

result, KANT can achieve better

translations than general-purpose

translation systems when applied in

specific technical domains.





Does any one have more info about it or has actually seen this software at work?



Thanks in advance,



ME



The same concept exist with the classic machine translation system such as systrans, though they don\'t publicize much about it to the general public.



A customized version is way beyond the prices of the desktop versions you usually hear about.



Basically, machine translation rely on terminology databases and rules to analyse the original text, to find \"appropriate\" equivalent terms in the target language and to organize that terminology



Individual terms choices have a higher or lesser usage index allowing the software to select a proper term.



Now, as you know what your documents are about, you can impose a software such as KANT to use, and give preference to, the terms of a specialized glossary.



So obviously, the output of the machine translation is improved compared to a machine translation using irrelevant terminology databases.



KANT was coming together with an other initiative, controlled writing, called CTE, Caterpillar Technical English. The idea was to write in a simplified, more structured English (The more you follow rules, the more chances there will be for a software based on those rules to match them and provide a sensible \"translation\") with a unified terminology extracted from past Caterpillar documentation.



All that terminology was translated in the Target languages (AMT) and input in KANT. technical writers were trained to write in CTE.



The result was then to be fed to KANT then given a review by translators human translator specifically trained in AMT.



The goal was not to create good natural translations, but generate a somewhat sensible result needing just a \"touch and a lick\" from the proofreader to convey the technical meaning, instead of a complete rewriting. The final result was not to be a natural language output (which is the purpose of human translation) either, but a technical lingo that would be understandable to all users and consistent throughout the 50 millions + words of Caterpillar documentation.



That was before 98. I don\'t really have recent stuff on that, but I worked on some Caterpillar projects last year, and it was translation, not proofreading, so I suspect the project did not accomplish it\'s goal yet. However, it must be said they have a hell of a terminology database!



I would very much welcome more recent informations. The idea was a very interesting one, but I suspect they underestimated the amount of effort needed to create a new lingo with a fixed structure, classify terminology, define the terms(that\'s a bit of a missing part), teach human authors to write for a software in a way that can be understood by an user, program the said software and relay that into the other languages.



A great idea in fact, but it\'s a hell of a work. Really. And the bulk of that work concentrate on a single subject, meaning that if each technical field engaged in a similar pursuit, the largest part of that work would have to be done all over again for each subject, each brand.



If that\'s what you were wondering, have no fear, they ain\'t gonna replace us this decade, and probably not the next.

Direct link Reply with quote
 

Maria Eugenia Farre  Identity Verified
Brazil
Local time: 20:16
English to Portuguese
+ ...
TOPIC STARTER
No such worries yet Mar 20, 2003

Quote:


If that\'s what you were wondering, have no fear, they ain\'t gonna replace us this decade, and probably not the next.





Hi Sylver,



Thanks for your comments. I was just surprised to see that Caterpillar was going the other way, developing automated translation instead of TM tools. We\'ve all been doctrined to believe that automated translation is no good and that no good will ever come of it.



CTE seems to be a great concept by the way. Is it still in place at Caterpillar? I seem to have read somewhere that Caterpillar relies heavily in Deja-vu today, although they go the whole range of CAT tools according to this article

http://www.internationalwriters.com/dejavu/Integrating_tools.html



ME

Direct link Reply with quote
 

sylver  Identity Verified
Local time: 07:16
English to French
Machine translation- A dream that won't come true ...fully Mar 20, 2003

For CTE, take a look at www.tc-forum.org/topiccl/cl15diff.htm Worth a few minutes of your time.



Another great reference is www.controlledenglish.com/history.html. That web site deserves a visit.



I wouldn\'t be able to say wether or not Caterpillar evolution of CTE is turning out to be profitable, but I am inclined to think so as the use of CFE and CTE made their translators\' work a lot more consistent, and easier too. (see exemples of the link above)



CTE, or the idea, is still worked on. A hell of a job too, but as far as technique is concerned, that\'s the way to go, IMVVVHO. Technical subjects are developping a new lingo. We might as well make it universal, clear and easy to use right from the start. 6000 languages is already way too much for such a small planet. Hell why can\'t everybody speak French, just like sensible persons do!



Afaik, Machine translation has been and is still dreamed about and worked on. In this interesting link you provide, you can see that Déjà vu is using MT in conjonction with it\'s TM\'s matches to improve it\'s fuzzy matches. That\'s an other way to say that they are taking a peek at machine translations, isn\'t it?



Based on the cheer volume of documents that do require translation, a solution will have to come out to simplify the process, somehow.



Computers do not think, so machine translations (real - implying understanding) can not exist and will not. However, through CL and codification of terminology, a certain type of MT may reach a level of usuability. It\'s pretty close. The subject of AI (MT relies on AI) as it is currently presented, aims at faking intelligence. The same is true of MT, but while *faking* ain\'t good enough to qualify for AI, faking translation could do very well to support our work.





Direct link Reply with quote
 


To report site rules violations or get help, contact a site moderator:

Moderator(s) of this forum
Maria Castro[Call to this topic]

You can also contact site staff by submitting a support request »

KANT Translation System- Automatic Translation at Caterpillar

Advanced search


Translation news





WordFinder
The words you want Anywhere, Anytime

WordFinder is the market's fastest and easiest way of finding the right word, term, translation or synonym in one or more dictionaries. In our assortment you can choose among more than 120 dictionaries in 15 languages from leading publishers.

More info »
Across v6.3
Translation Toolkit and Sales Potential under One Roof

Apart from features that enable you to translate more efficiently, the new Across Translator Edition v6.3 comprises your crossMarket membership. The new online network for Across users assists you in exploring new sales potential and generating revenue.

More info »



All of ProZ.com
  • All of ProZ.com
  • Term search
  • Jobs