RIF kickoff f2f breakout session: classification of rule languages
Note: This page is intended to be a record of the breakout session, not an evolving document. Do not modify this page except to make corrections to the record.
San Francisco, 9th of December 2005
chair: Chris Welty
scribe: Pascal Hitzler
Welty:
- I imagine clusters within which interoperability is good and desirable\\
- between clusters interop will be imperfect or "lossy".
- thats for phase 2
- phase 1 only horn, so probably no problem
Hitzler, Grosof:
- phase 1 probably not so easy
classification dimensions brainstorming
"Market Segments" (Minutes clarification - Donald Chapin 2006-01-17: The concept discussed was "types of rule systems that are in use in the marketplace")
- Relational DBMS
- PR-SR
- ELA
- Prolog
- controlled english LP i.e. "rules systems supporting business communication of rules"
(Minutes clarification - Donald Chapin 2006-01-17: This was my suggested rule system type. LP stands for Natural Language Processing. The terminology I used to name this category of rule system was "Rules Systems Supporting Business Communication of Rules". A description of this category of rules systems can be found on Wiki page: Rules Systems Supporting Business Communication of Rules)
"Expressiveness"
- FOL vs. Modality
- Decidability
- vs. uncertainty
- vs. Logic programming (LP)
"Semantic assumptions"
- CWA (closed world assumption) vs. OWA (open world assumption)
"Support Features/Implementation of system/Meta-Data"
- inference control:
- backward vs. forward
- incremental vs. exhaustive
- authoring
- analysis/testing
- comments
opinions (not necessarily chronological) raised while brainstorming
will priorities etc. play a role?
Kifer:
- syntactic vs. semantic classifications?
Barkmeyer:
- CWA/OWA distinction may be difficult
Chapin:
- what about linguistic logic?
- has to do with grammar as in linguistics
- may be market segment
- e.g. controlled english LP
Hitzler:
- need OWL ontology to classify rule systems
- (wide agreement on this)
- Procedural attachments?
Welty:
- Is feature, shouldn't be used in classification.
- Are these support features important for us?
Grosof:
- secondary at this point, but still important.
Welty:
- Its metadata.
Engel:
- What's should be the result of the WG? Joint syntax?
Grosof:
- Don't quite know. Probably difficult.
Kifer:
- probably common syntax not even possible, e.g. different logical connectives
Welty:
- alternative is one syntax construct with annotations which tell which
- semantics is meant
Welty/Kifer:
- clear that we will have different semantics
Menzel:
- probably better focus on something small rather than branch out as
- we do now, i.e. build RIF bottom-up in an incremental way?
Grosof on historic approaches:
- SWRL: LP syntax (with FOL semantics) on top of OWL SWSL: Expressiveness of both LP and FOL at disposal
Chapin:
- Why not use FOL/LP distinction as start for design of RIF?
Welty:
- building OWL ontology of rule languages.
- which properties?
- which key differences (discriminators) to split things up? (see below)
Kifer:
- should have semantics vs. syntax table.
Welty:
- That can be one set of discriminators, but not the only
towards an OWL rules ontology
discriminators:
- open vs. closed
- FOL vs. other semantics
- non-monotonic?
- forward/backward chaining
- procedural vs. declarative
- decidability
- complexity
- modality/intentionality
Welty:
- lets start with this list and elaborate on it
- calls for suggesting an ontology
Menzel:
- reuse some other similar work already done?
Hitzler sends notes to Welty.
Everybody proposes taxonomies.
Set up Wiki page.
Hitzler sets up the Wiki page: Rulesystem Arrangement Framework.
[End of session/Welty has to leave]
post-discussion:
Grosof:
- can we work bottom-up in parallel?
- proposes Horn plus Lloyd-Topor
Chapin:
- what about using common logic? SBVR is fragment of it
Grosof:
- doesn't like some non-FOL features of common logic. could also use FOL RuleML