OWLED 2007

From LeibnizWiki
Jump to: navigation, search

Contents

Day 1

Session 1 - Industry Track

Near Real-time Enterprise System Integretion - Suzette Stautenburg

Use case:

  • Very easily extensible tracking of e.g. sattelites, enemy forces, etc.

Wants:

  • Uncertainty
  • N-ary predicates
  • Individual denoting functions

Lockheed-Martin - Henson Graves

Use case:

  • integrate, organise, identify enterprise data
  • Assessment of requirements, design analysis

Wants:

  • Ontology-driven data insertion: using fill-in forms based on constraints on the ontology

IBM Watson - Anand Raganathan

Use case:

  • Stream processing of data streams (e.g. traffic camera data)
  • Need workflow description

Has:

  • System for describing workflows different from OWL(S)/WSML
  • Instance-level graph patterns
    • More expressive -> cycles, xsd sequences, progagate information using variables
    • ABox assertions instead of TBox axioms
    • Planner for automatic composition of optimal workflows

Renault - Francois Servant

Use case:

  • Diagnostics of car parts using RDF/OWL Full
  • Use their own reasoning tools
  • Each test has a cost:
    • Disassembly graph of parts in a car
    • Disassembly state
  • OWL Modeling is not "extra" work
  • Allows a lot of business knowledge about data
  • Easily implemented by using the "dynamic object model"-pattern in Java

Problem:

  • Still need to produce the data

Questions

Bijan: What part of OWLFull did you use? Would punning be enough?

Peter: Anyone doing any *OWL-Fullishness*??

Henson G: We use XML proof checking, decdability is not important

Suzette: Would like an integrated suite, currently using XML tools

Anand: Uses Protege 3.2beta, and uses an IBM-developed DLP reasoner

Francois: Uses Protege as well, would like a way to insert instances in a browser

Anand: Reache & replace over large ontologies, simple text editing.

Suzette: OWL/SWRL 2 Prolog -> as they use Prolog for reasoning.

Session 2

Brian Lowe

VIVO Website (Virtual Life Sciences Library), ontology based, at: http://vivo.cornell.edu

Based on java/tomcat/lucene/mysql

RDFS Semantics is sufficient for most cases, but need:

  • How to hide stuff in a database
  • Transitive properties

-> Moving to Jena/Pellet seems promising. Could use rapid/dynamic TBox editing for site editors.

Ontology of Fine Rolls (Henry III, 1216-1248) - Jose Viera

The ontology is used for describing person, subject, place etc.

Extract data from XML to ontology -> show in the portal

Problem: synchronisation between XML serialisation and RDF/OWL representation

Stephen Larson

Ontology that integrates (image) data accross diferent scales (microscopic, brain tissue), but also other sources.

Use *Algernon* for rules. Used to generate inferred instances from existing ones -> e.g. generate a neuron if you know there's an axon.

SWRL Tab - Martin O'Connor (SMI)

Use case: clinical trials application & SWRL

  • Create some consistency accross different trials

Need/Does:

  • Ontology Mapping (EPOCH/Bridge)
  • Knowledge-level queries
    • SWRL as query language (SPARQL is RDF, SWRL is OWL)
  • TBox, ABox queries, temporal queries
  • Implemented the SWRL Query Tab (JESS): SQWRL
    • Allows OWL restrictions in queries, disjunction and negation

Session 3

Integrity Constraints - Boris Motik

Integrity constraints (as used in relational databases) can be used for error detection:

  • Domain constraint
  • Participation constraint

Constraints can be used to find lacking information.

Axioms should be treated as a *check* -> current OWL reasoning is actually unnecessarily complex

  • Strategy is to split the TBox
    • Normal TBox - for inference rules
    • Constraint TBox - integrity constraints

If IC's are ok, we can throw them away for query answering.

  1. Model ontology as usual
  2. Mark (some/all) axioms as integrity constraints
  3. Add individuals
  4. Check IC satisfaction
  5. Forget IC's
  6. Answer ABox queries -> reason more efficiently

This procedure will lead to:

  • Better modelling
  • Better performance

Current issues:

  • Still needs to be implemented (will be in KAON2)
  • Computational complexity is unclear

Fuzzy Description Logics - Giorgos Stoilos

Implemented:

  • Fuzzy reasoning engine, FIRE (connected to triples in Sesame)
  • Fuzzy DL-Lite (fuzzy cardinality constraints)
  • Fuzzy DL

Can OWL model Football Leagues - Diego Calvanese

  • Need identifiers -> normal in ERM, databases etc. the notion of *key*
  • Identifiers exist in a restricted form in DL (the at_most_1 construct)
  • Developed DL-LiteA, which has the same complexity as relational databases.

SPARQL-DL - Bijan Parsia

  • Many query languages: DIG, nRQL, OWL-QL (OQL), SPARQL, SeRQL
  • DIG is not satisfactory for OWL-DL, many other query languages aren't either
  • Queries are the way in which users interact with the store.
  • SPARQL-DL is an extension of SPARQL which allows to use OWL-DLlish constructs (subClassOf, equivalentClass etc.) in SPARQL queries.

Panel 1 - SPARQL/OWL/Rules

  • What's the difference between Queries and Rules?
  • Should it be OWL+rules or owl+RULES

Sandro Hawke (W3C)

RIF is a WG at W3C

  • Core+extensions
  • Interchange (not native!)
  • May be lossy if you go beyond core
  • Uses Horn logic
  • Two way;s to use OWL on RIF-only systems
    • Hybrid systems likely for OWL+RIF users
  • Looking for the intersection

Martin O'Connor

  • SPARQL is grotesque: poor syntax, unclear semantics, no extensions
  • Exposes too much of RDF to the user
  • SWRL has clean syntax, clear semantics
    • supports extensions through plugins

Boris Motik

  • We *should* distinguish Queries from Rules
    • Query semantics is not controversial
    • Rules have different focus: facts. Furthermore, their semantics *is* controversial
  • Extend SPARQL, and fix it. Map it onto full algebra
  • Add negation and universal quantification
  • Make semantics on OWL unambiguous
  • for ABox reasoning:
    • Rules
    • Integrity constraints

OWL+Rules is easier if we first extend it with integrity constraints

Italian Guy

  • KR Perspective
  • Overcome undecidability of QL's by using auto-epistemics

Bijan Parsia

  • After a survey of submissions, and emails sent, mentions as primary 'wishes':
    • SWRL, DL-Safe, Scalable ABox, Constraints, Keys, SPARQL-DL
    • "SWRL" could mean anything: rules, constraints & queries

Questions

Alan Ruttenberg:

  • Have the feeling that the "room" wants mixed ABox/TBox queries

Alan Rector:

  • We need a list of typical rules and checks for rules formalism expressivity (RIFRAF was a start in the RIF WG)

Ulrike Sattler:

  • How do you know where to draw the line between Rules & OWL?

Suzette:

  • As much as possible in OWL

O'Connor:

  • But people just use the most convenient method.

Session - OWL 1.1 Status Report - Ian Horrocks

  • Invitation to do document-editing

Sandro Hawke will be staff-contact for possible WG:

  • WG usually runs for 3 years
  • 15 people
  • F2F meeting 4 times per year
  • Phone conference every week
  • 1-2 days per week per person
  • Proposed chairs: Ian Horrocks & Alan Ruttenberg
  • Call for participation at the start of October, first telecon at the end of October
  • F2F at the end of November, first week of December.
  • Talk to AC Rep to speak-up for the initiative!

For next OWLED:

  1. Report of the possible
  2. Proposal
  3. Specification

Session

OWL API - Matthew Horridge

The WonderWeb API has transformed into the OWL API:

  • Syntax neutral
  • Closely follows specification
  • change-objects, easy redo/undo
  • Multiple ontologies
  • Many syntaxes, can be connected to OWL/RDF triplestore
Personal tools