DARPA Agent Markup Language (DAML) Details

Mike Dean
Principal Investigator, DAML Integration and Transition
[email protected]

Semantic Web Panel
Intelink Conference
St. Louis, MO
29 August 2001

http://www.daml.org/2001/08/intelink-panel-mdean/Overview.html

$Id: all.htm,v 1.5 2001/08/29 17:08:29 mdean Exp $

Plan

discuss DAML+OIL language and tools at a deeper level of detail
- demo and discussion of DAML tools based on (portions of) the Intelink Conference home page
guided tour of http://www.daml.org

Baseball Ontology

last weekend I started thinking about a baseball ontology as a training tool
on the plane coming here, I started developing it to test a new version of a DAML tool
we'll use it to express
```
St. Louis Cardinals vs. San Diego Padres - August 28, 29, & 30th
```
from the Intelink Conference home page on the Semantic Web
steps
1. look for an existing "baseball" or "sports" ontologies in the DAML Ontology Library
2. finding none, build a new ontology using OilEd
  - tool developed by our EU Semantic Web collaborators
3. review the ontology using dumpont
4. create instances (small scale) using OntoMat
5. check the resulting object graph (GIF) (SVG)
6. report results using WebScripter
7. download (selected) information to PalmDAML
8. submit resulting ontology to DAML Ontology Library
results are at http://www.daml.org/2001/08/baseball/

Large Scale

image if the daily Sports Section were published in DAML
- with a subscription, over time you too could answer questions like "who has the highest batting average against left-handed pitchers on Tuesdays in July"
"all" of this data already exists somewhere
- want to provide access to databases rather than duplicating huge amounts of data
- DAML Services to identify, describe, locate, and access such databases
- DAML Query Language(s) to access such databases
- mapping of various existing "ontologies" and content to DAML
  - UML
  - XML Schema
  - ERwin
  - Horus Data Access Toolkit
  - ...
real owners should control data (Major League Baseball, individual teams, Hall of Fame, etc.)
what's the business model?

Technical Issues

"ontology creep"
- baseball ontology is already getting big
  - would it be better to split into a family of ontologies: teams, games, etc.?
  - could/should we extend the ontology to support football, soccer, etc.?
aggregation
- computing box scores from individual pitches, hits, etc.
- computing standings from individual games

Why DAML?

(vs. database, base XML, etc.)

URI-based naming
- "Cardinals" vs.
  - http://www.mlb.com/teams/2001#cardinals
  - http://www.audubon.org/birds#cardinal
  - http://www.stanford.edu/athletics#cardinal (color)
- drill down
adding properties to someone else's objects
- Goodyear Blimp pilot needs lat/lon for Busch Stadium rather than street address
- aggregation

expressive power

class expressions

<daml:Class rdf:ID="Southpaw">
  <daml:sameClassAs>
    <daml:Restriction>
      <daml:onProperty rdf:resource="#position">
      <daml:hasValue rdf:resource="#Pitcher">
    </daml:Restriction>
    <daml:Restriction>
      <daml:onProperty rdf:resource="#throws">
      <daml:hasValue rdf:resource="#Left">
    </daml:Restriction>
  <daml:sameClassAs>
</daml:Class>

subproperties

<rdf:Property rdf:ID="father">
  <rdf:subPropertyOf rdf:resource="#parent"/>
<rdf:Property>

data distributed over multiple sites
fine-grained filtering of individual facts rather than pages

www.daml.org Guided Tour

More Information

Horus demo at NIMA tomorrow
http://www.daml.org/ (start with roadmap)