call notes: ---why, formula interning, ...

* DanCon takes a call from timbl...
<DanCon> TimBL: I put in the --why stuff; it seems to work for DAML...
<DanCon> ... and I wrote check.py
<DanCon> ... then I tried it on the ACL case...
<DanCon> ... ran into a problem; it's a general one: I intern the
formulas. if you conclude the same thing for different reasons...
<DanCon> DanC: I was gonna ask you about that...
<DanCon> DanC: how about a why[fmla] -> justification table?
<DanCon> er... justifications()
<DanCon> er... justification(s)
<DanCon> DanC: doesn't it intern formulas when it hits } ?
<DanCon> TimBL: hmm... good idea; might save some tricky recursion
<DanCon> ---
<DanCon> TimBL: I wrote a diff.py which does daml-level smushing
[equality reasoning]
* DanCon finds $Id: diff.py,v 1.1 2002/12/04 15:21:44 timbl Exp $
<DanCon> TimBL: we talked about if-added rules; I could make that an
if-added rule...
<DanCon> ... for equality, one normalized form is, for X=Y where X lex<
Y, store properties on X, just note that X=Y somewhere.
<DanCon> ... this is optimal for lots of things, ... queries...
<DanCon> DanC: actually, for queries, the easiest thing is to have all
the properties on all terms...
<DanCon> DanC: I wonder if backward chaining is better for equality
stuff... maybe not... circles seem to come easily
<DanCon> --
<DanCon> TimBL: I added an explain() method to formulas... not sure I'm
happy about that...

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Monday, 9 December 2002 18:21:48 UTC