At 11:16 +0100 24/06/03, [EMAIL PROTECTED] wrote:
Dear all,
I was wondering if someone knows of any specific citations in publications or any article(s) that discourage MT for High Quality translation efforts, and/or those that specifically promote MT systems only for dissemination purposes using technical manuals.
There are such articles with these 2 points of view. But both are too naive.
HQMT can be achieved and beat HT in some contexts. That means that postedition of raw MT costs less thant postedition of HT. revision can be counted in minutes/page (20 after "standard" HT) or number of editing operations per 100 words (20-25 after stantard HT), if counting insertions and suppressions of entire words (replacement costs 2, exchange 2, swapping around a segment 4). See reports on Meteo (3% revision), CATALYST (CMU-Caterpillar), ALT/Flash, even METAL (Slocum 1984). HQMT can also be achieved with human intervention during or after analysis, with a view to multitarget translation. See Herv� Blanchon's thesis on all such "interactive systems", plus a paper in MT (1994-95). You may again evaluate quality by the human time taken to obtain N polished translations (maybe 15 minutes/page of interactive disambiguation and 5mn/p of postedition instead of 20, for each target language).
To advocate MT only for dissemination of technical manuals is plain silly when considering the millions of pages translated by MT web servers everyday on request of internauts wanting to access to information.
Please keep in mind the figures from the evaluation of the GAT (Georgetown Automatic Translation) system at Euratom, Ispra, around 1972: linguistic quality 1/10, usefulness 9/10. They still apply�
Regards,
Lorena Guerra, Postgraduate student at DCU- Ireland. Information needed for my dissertation in Translation Technology.
Best regards, Ch.Boitet
-- For MT-List info, see http://www.eamt.org/mt-list.html
