PAW . cwm dev

CWM developer/PAW

23 Nov 2005


eikeon/DK, vlatko/VK, DanC/DC, timbl


  1. Convene, take roll, review agenda
  2. Around the table
  3. comparing formulas with variables
  4. term API in javascript, eulermoz
  5. GRDDL support in cwm; xslt engines
  6. PAW demo: proof gen/checking
  7. cwm/pychinko integration
  8. Term API
  9. next meeting



Convene, take roll, review agenda

<timbl> I had kinda omitted to think of the meeting I'm afraid, no agenda.

<timbl> We could discuss I suppose papers, and changes.

<eikeon> dwood and I chatted about maybe adding a quick demo item to one of the cwm/paw meeting agendas if that works... since we've been the only two on the paw/demo calls.

<timbl> Ok, we'll do a quick ad-hoc agenda

<scribe> Agenda: v 1.75


Around the table

DK: getting rdflib into release state for use elsewhere in the lab...
... not much progress on Term APIS

VK: spent the week mostly doing pychinko hacking. log:conclusion and log:conjunction added minimally working version
... I do log:conclusion with a sub-rete.


VK: worked OK because the rete is a 1st class object. [yay! no globals!]

<DanC_> r5336 | kolovski | 2005-11-21 21:19:57 -0600 (Mon, 21 Nov 2005) | 2 lines

<DanC_> -note this is a minimally working version

<vlatko> yes

TimBL: haven't done anything directly relevant, but did think about the term API in javascript...

DanC: I was at XML2005 all last week

<DanC_> RDF Calendar, GRDDL, Microformats, and all that at XML2005 in Atlanta

DanC: while working on figures for slides, had to use an old version of cwm cuz -bySubject isn't tested and, of course, rotted

<DanC_> cwm --bySubject breaks bnode refs Dan Connolly (Friday, 11 November)

timbl: hmm... yeah... not all the -flags are tested well

<timbl> created swap/test/output directory for such tests

comparing formulas with variables

VK: the rete is efficient because we use dictionaries: pred/obj/sub
... matching formulas is tricky [not sure I'm following, but clearly timbl is]

TimBL: ... unify ...
... compare two formulas...

VK: I tried that, and it works great with no variables, but with variables...

<vlatko> fails:

<vlatko> {{ :a :b :c} ?y {:a :b ?x}} log:implies {<test> a <SUCCESS> }.

<vlatko> {:a :b :c} :h {:a :b :c}.

<vlatko> succeeds:

<vlatko> {{ :a :b :c} :h {:a :b :c}} log:implies {<test> a <SUCCESS> }.

<vlatko> {:a :b :c} :h {:a :b :c}.

TimBL: take care with scoping of ?foo syntax

DC: you might try just explicitly quantifying them at the top level

<timbl> <test> a <SUCCESS> .

<timbl> @forAll :x, :y.

<timbl> {{ :a :b :c} :y {:a :b :x}} => {<test> a <SUCCESS> }.

<timbl> {:a :b :c} :h {:a :b :c}.

<vlatko> @forAll :x,:y.

<vlatko> {{ :a :b :c} :y {:a :b :x}} log:implies {<test> a <SUCCESS> }.

<vlatko> {:a :b :c} :h {:a :b :c}.

@forall ?y. {@forall ?x. { :a :b :c} ?y {:a :b ?x}} log:implies {<test> a <SUCCESS> }

DC: vlad, do you handle scopes other than global yet?

VK: dunno.. is there a test?

TimBL: rules13, I think...


term API in javascript, eulermoz

TimBL: played with the term API in javascript...
... worked with Jim Ley's RDF parser...




[there's an eulermoz... I wonder how it relates]


hm... empty




"We please to announce we commit N3 scanner first version."

<timbl> Age: 14 months


GRDDL support in cwm; xslt engines

DanC: I almost put GRDDL support in cwm

TimBL: depends on XSLT.... hmm...

DanC: 4suite might have one, but I'd use libxslt, with a --grddl or --xslt flag, ala the way crypto stuff crosses the pure-python barrier

<DanC_> RFE: GRDDL support 09 Jun 2004

PAW demo: proof gen/checking

<timbl> We are still I think at the point that the cwm --why doesn't work on DanC's machine, doe son Vlad's and Tim's

<DanC_> yeah; I haven't summonned the energy to chase after that heisenbug

DK: wondering what to do next... we're using a fake proof so far..

DC: since the proof check doesn't work on my desk, maybe you could do the next step? I sent mail about where I was stuck:

Subject: proxy stuck: trouble figuring out cwm proof gen API

Date: Fri, 28 Oct 2005 17:32:12 -0500

Subject: Re: proxy stuck: trouble figuring out cwm proof gen API

Date: Thu, 03 Nov 2005 16:57:58 -0600

Committed revision 160.

TimBL: I'd like to track that down, using more and more debug output

<DanC_> (I'd rather turn all the debug output into unit tests... hmm...)

<timbl> if it is a particular combination of features which breaks then unit tests won't find it

DK: ok, I see the mail; I think I can take the next step... I'm not sure how to check the proof...

<vlatko> for checking:

<vlatko> python ~/w3ccvs/WWW/2000/10/swap/ -v50 -c40 ,pf.n3 2>,pfcheck

<vlatko> generating a proof:

<vlatko> python --test-rein 2>,errs >,pf.n3

VK: I've run the checking stuff. it works for me

<timbl> check -v50 < ,pf.n3 > result.n3

cwm/pychinko integration

DC: vlad, are you working on using the rete inside cwm?

VK: thinking about it...
... was thinking about static analysis to see if rete can handle it

DC: how about just a --use-rete flag and crap out of it won't work?

VK: ok, that's even easier

TBL: I think yosi's latest work starts with a KB, looks for rules, and builds a rete

DK: yeah

VK: and then put it back?

TBL: i think conclusions from the rete are handled just like cwm: either back in the KB or in a new formula [?]

VK: maybe I can contact yosi about it

Term API

DK: anything on the term api/interface?

<DanC_> re term API, v 1.5 2005/11/17 15:42:10

<DanC_> cwm API notes:

<DanC_> Symbol.uriref() vs uriref2()... why not an optional arg?

<DanC_> -- convert RDF to iCalendar syntax

DK: I've been tracking

DC: ok, I can scribble some article ideas there

<eikeon> We should be able to converge on API enough to interop on scripts such as -- can start by comparing the differences in the API the scripts expose.

next meeting

regular schedule calls for 7 Dec

DanC on a plane back from TAG meeting about that the time

TimBL: I'm ok for 7 Dec 14:30ET

<vlatko> I'm ok, too

VK: I think I'm available, though end-of-semester foo creates risks
... 27 Dec I take holiday

<vlatko> till 17th Jan.,

<vlatko> .


Minutes formatted by David Booth's scribe.perl version 1.127