From: pat hayes (phayes@ai.uwf.edu)
Date: 07/18/01
>Hi folks,
>
>With regard to what kind of inferencing is reasonable to consider to
>support in DAML, in combination with DAML+OIL, as an inference (as
>opposed to transformational) rule language/expressive-extension:
>
>I believe that the starting point should be declarative logic programs (LP),
>first without negation ("hornlog LP"),
>then with negation-as-failure ("ordinary LP"),
>then with prioritized conflict handling ("courteous LP") and
>procedural attachments ("situated courteous LP").
I find this odd, since this sequence seems more like a
transformational rule system than an inference rule system. First,
negation-as-failure is invalid in DAML+OIL, so the second step seems
odd for an inference rule scheme. (And BTW, this was not an arbitrary
decision, but seemed forced upon us by the necessities of Web usage,
which require a monotonic logic.) Second, presumably 'procedural
attachements' could be sued to perform any transformations (?), so
the end result is clearly a transformational rule system rather than
an inference rule system.
It is obvious that LP as a methodology is very useful, but rather
than the basic web intercommunication language being made
nonmonotonic, I would prefer to see a monotonic rule system which can
accept as input an assertion that a certain namespace is closed, so
that it becomes valid to use negation-as-failure, and hence ordinary
LP, *in that namespace*. Let me call this "localized LP". What we
need, therefore, is a way to make the required assertions.
>The first step in Webizing this is to permit the logical constants
>(relation symbols, function/constructor symbols) to be URI's.
>The second step in Webizing this is to incorporate the DAML+OIL features.
>On the way, RDF as a form of essentially syntax, well as XML as a more
>usual kind of syntax, should be supported.
>
>The latest serious attempt at defining a standard for all this is
>RuleML (http://www.dfki.de/ruleml, also see
>http://www.mit.edu/~bgrosof/#XMLRules), for which a new version (0.8)
>of DTD's has just this week been released. There are several
>research-world tools now available for it, including to execute a
>version of it in the XSB logic programming system. IBM is planning to
>support it in CommonRules (which already supports situated courteous
>LP and translation to/fro XSB and KIF
Do not put too much effort into KIF until we get the new standard
written. ISO-KIF will be considerably easier to interface to than
current KIF (UTF-8 character codings, simpler syntax, coherent
semantics, much more flexible & expressive notation, better treatment
of sequence quantifiers.)
>, with its own earlier XML format
>BRML).
>
>RuleML has not yet tackled incorporating DAML+OIL features, so that is
>one of the next things to do. Please contact me if you'd like to work
>together on that.
I would like to work together, given the above remarks, yes, if you
are interested.
Pat
---------------------------------------------------------------------
(650)859 6569 w
(650)494 3973 h (until September)
phayes@ai.uwf.edu
http://www.coginst.uwf.edu/~phayes
This archive was generated by hypermail 2.1.4 : 04/02/02 EST