- From: Nishant Shukla <shukla@dm.ai>
- Date: Tue, 4 Feb 2020 08:27:14 -0800
- To: Dave Raggett <dsr@w3.org>
- Cc: public-cogai@w3.org
- Message-ID: <CAAQqOn8eGaTEToUddtf0eQrNgmSwTcEW2SJOZSp5bEgC3X4KDA@mail.gmail.com>
Thanks Dave, That sounds like a great idea. A vendor/client relationship has some well-defined communication patterns, and lends itself to be modelled algorithmically. Speaking of modelling dialogue, I've studied a concrete example of a chatbot that sells cakes [1] by Dr. Morbini (back when he worked at the University of Southern California). I've rewritten some of the dialogue in DMScript [2], and can further convert DMScript to chunks easily, since both share the concept of chaining rules to resolve a goal. I'm considering writing that cake-vendor scenario using chunks for the NLD demo in the GitHub repo. What do you think? [1] https://github.com/fmorbini/jmNL/blob/846dfebc339ea5b30487ef483fe50c8b151a0b05/resources/characters/CakeVendor/dm/textFormat/policy.txt [2] https://github.com/conversational-interfaces/dms-scratchpad/blob/master/examples/cakevendor.dms On Fri, Jan 31, 2020 at 12:05 AM Dave Raggett <dsr@w3.org> wrote: > Hi Nishant, > > Your help would be welcome in respect to work on natural language > dialogues. I am proposing to first work on a demo where the language usage > and meaning is clear and well understood, in this case for the dialogue > involved in ordering a meal at a restaurant. This will allow us to work on > representing the meaning and the semantic processing needed to support the > dialogue, including disambiguation of word senses, and finding the most > appropriate bindings for prepositions, pronouns etc. > > Computational linguists have emphasised linguistic processing over > semantics, but with a cognitive model we can redress the balance, and > simplify the lexicon accordingly, along with reducing the amount of > statistical data commonly associated with statistical parsing. You can see > more at: > > https://github.com/w3c/cogai/blob/master/demos/nld/README.md > > Note that further demos would look at how cognitive agents can mimic human > abilities to learn new language patterns from small numbers of examples. > > Best regards, > > Dave Raggett <dsr@w3.org> http://www.w3.org/People/Raggett > W3C Data Activity Lead & W3C champion for the Web of things > > > > -- Nishant Shukla DMAI, Inc. 707-574-8552 | Schedule <https://calendar.google.com/calendar/embed?src=shukla%40dm.ai&ctz=America%2FLos_Angeles>
Received on Tuesday, 4 February 2020 16:27:47 UTC