From: Deborah McGuinness (dlm@ksl.stanford.edu)
Date: 08/19/02
thx for the agenda.. The topic of justifications came up as possibly being included or at least pointed to in what we put out with respect to dql. We have done some work on a specification of a justification in daml+oil that we are currently testing internally for use with our reasoner and we are in the evaluation stage for reasoners used in shaken (one of the two teams in the rapid knowledge formation program). it is along the lines of the explanation work we did with an early dl system (classic) and later a tableaux DL system and then also we used the same line of thought for our early explanation module for atp (a model elimination theorem prover at stanford). Now our thought is to provide the daml+oil (soon to be owl) specification so that reasoners can pass proofs along with their conclusions to collaborating agents. essentially one can get a justification for any piece of information that the system contains. justifications include the last inference rule used (with bindings in the antecedent and conclusion of the inference rule). inference rules can be thought of in natural deduction style format with a inferenceRuleName, 0 or more antecedents, and exactly 1 conclusion. so asking why conclusion xx is believed will yield a inference rule name (with associated bindings) from which a reasoner can conclude xxx. followup questions are generated by asking why one or more of the antecedents of that inference rule is believed. in the most complete implementation we did of this approach (in classic) if conclusions could be deduced by more than one last application of an inference rule, then there were multiple explanations available. hopefully in a week but probably more like 2, we will have an implementation that is standalone in that it does not require our reasoner (jtp) to be used. we just started a new postdoc on this work and would be interested if anyone wanted to be an early alpha evaluator/user. I expect him to have a little ramp up time but then i expect to be making significant progress. i bring up this topic for 2 reasons: 1 - I am interested in knowing if anyone is interested in being an early adopter/evaluator 2 - I can add some level of detail of this work to fill out the dql literature on the portion that says the form of justification is not specified yet. I could add something saying a justification could take the following form.... and then give pointers to the work in whatever detail we think is appropriate. deborah Mike Dean wrote: > We'll have our weekly telecon tomorrow (August 20) from > 4-5pm EDT, 1-2pm PDT, and corresponding times in other > locations. > > The dial-in numbers are > > U.S.: 866-779-0774 > > outside U.S.: +1-334-309-0263 > > After dialing, enter *7289913* (including stars) > > If you're asked, the chairperson is Mike Dean > > A parallel logged IRC session will be available on the > #joint-committee channel at irc.daml.org (see > https://www.daml.org/irc/ for the server password). > > Below is a draft agenda; additions are welcome. > > Mike > > agenda updates/additions > > announcements > > minutes deferred > > DQL updates (Ian, Pat, Richard, et al - 10 minutes) > > DQL decision to publish (all - 10 minutes) > > possible DQL papers (Richard, et al - 10 minutes) > > DAML Rules kickoff (Benjamin, Ian - 10 minutes) > > DAML Rules working session (all - 20 minutes) > > next week -- Deborah L. McGuinness Knowledge Systems Laboratory Gates Computer Science Building, 2A Room 241 Stanford University, Stanford, CA 94305-9020 email: dlm@ksl.stanford.edu URL: http://ksl.stanford.edu/people/dlm (voice) 650 723 9770 (stanford fax) 650 725 5850 (computer fax) 801 705 0941
This archive was generated by hypermail 2.1.4 : 08/19/02 EDT