Hi Vadim, Enjoyed reading about your NLP approach.. Our software attempts to track domains of knowledge from sentence to sentence, but I have to say it can be thrown off the scent by a red herring.. Regards, Terence
On Tuesday 08 September 2009 06:26:39 Le Khanh Hung wrote: > дорогой Vadim, > > It is interesting to know about your project. > Congratulations > You'd build not sonata, but a symphony > > Best regards, > Hung > ----- Original Message ----- > From: Vadim Berman > To: Le Khanh Hung ; [email protected] > Sent: Tuesday, September 08, 2009 11:10 > Subject: Re: [Mt-list] Not the approach to blame > > > Hi Hung, > > Actually, our system is doing exactly that: "supercharged" rules > describing use patterns, plus WordNet-based data organisation. > > It is more than MT, rather a generic analysis and transformation > system... > > Currently it is used to "understand" things, we're extending it for MT. > We started with a "tough nut" which is Chinese. Not FAHQMT, but looks not > bad. If it goes as planned, we'll try our luck next year in NIST. > > I'm attaching a few screenshots, and throwing a couple of URLs as well: > http://www.digitalsonata.com/download/carabaoNLP.pdf > http://www.digitalsonata.com/faq.aspx > > Best regards, > Vadim > > ----- Original Message ----- > From: Le Khanh Hung > To: [email protected] > Sent: Tuesday, September 08, 2009 1:31 PM > Subject: [Mt-list] Not the approach to blame > > > Dear all, > > We consider that the rule-based approach has the bright future. > > What the rule-based paradigm has, now? > > 1 - The elegant and powerful Chomsky's formalism (and not so elegant > some its extentions) 2 - The elegant Wordnet architecture > > For FAHQMT, we must link this two things together > The question is HOW? > > > > --------------------------------------------------------------------------- >- > > > _______________________________________________ > Mt-list mailing list _______________________________________________ Mt-list mailing list
