- From: Dave Lewis <dave.lewis@cs.tcd.ie>
- Date: Wed, 09 Jan 2013 01:28:31 +0000
- To: "Dr. David Filip" <David.Filip@ul.ie>
- CC: Yves Savourel <ysavourel@enlaso.com>, Felix Sasaki <fsasaki@w3.org>, tadej.stajner@ijs.si, public-multilingualweb-lt@w3.org
David, You are right, this sort of opportunistic MT use of disambiguation data would suit more the home spun/push button trained MT market, I'd presume the big guys like Google would have sucked DBpedia dry of all useful parallel text already. The other two use cases you describe are also very credible options, and probably more immediately understandable to localization folk so very worth pursuing. cheers, Dave On 08/01/2013 22:25, Dr. David Filip wrote: > Dave, says that the disambiguation is more useful for MT. I do not > entirely agree. > I do not see large MT providers like Google building support for this > info. On the other hand, SOLAS consumes the disambiguation info and > presents it to the human agent, as a color-coded span with linked > reference, eventually prose comment. > > People keep talking about importance of terminology management, > however terminology has always lower priority compared to direct > leveraging, as it is harder to put a financial benefit tag on it. > Getting marked terminology candidates via ITS encoded disambiguation > info is very attractive bootstrapping scenario for human translation.. > Apart from that using disambiguation as term candidates in a broder > term life cycle is an obvious backdoor for terminology (term), which > is good.
Received on Wednesday, 9 January 2013 01:28:57 UTC