At 12:59 -0500 30/03/04, Ariadna Font Llitjos wrote:
Hi Nic,
I'm actually working on how to use user feedback to directly and automatically improve a translation grammar.
I have a paper accepted in LREC04 that describes the tool I built to allow users to correct translations (the TCTool). It's already linked from my web-site: http://www-2.cs.cmu.edu/~aria/publications.html
And I'm currently writing a paper that focuses more on the back-end details, which I could send you as soon as it gets accepted to a conference, if you're interested. Let me know.
ari
----------------------------------------------------------------------------- Ariadna Font Llitjos Language Technologies Institute Carnegie Mellon University 4611 Newell Simon Hall 5000 Forbes Ave. Pittsburgh, PA 15213 office 1 412-268-3653 fax 1 412-268-6298
[EMAIL PROTECTED] http://www.cs.cmu.edu/~aria/
That is another way to attack the problem of improving MT quality output, but it is not of the sort envisaged by the original question. However, it supposes (in general) that the user is a professional translator knowing very well both languages, while the ID approach is more for non-professionals knowing the source language only.
I would like to mention that user feedback to automatically modify MT systems (learning approach) is not a new idea. Pr SU Keh Yih has incorporated it in his commercial MT systems at BDC at least since the late 80's.
Dr. Keh-Yih Su [kysu AT bdc.com.tw] Behavior Design Corporation 2F, No. 5, Industry E. Rd. IV Science-Based Industrial Park Hsinchu, Taiwan, ROC Tel: (O) 886-3-5770243 ext 300 Fax: 886-3-5770459
Best, CB
_______________________________________________ MT-List mailing list [EMAIL PROTECTED] http://www.computing.dcu.ie/mailman/listinfo/mt-list
