XML + Semantics = DARPA Agent Markup Language

Principals, Practice & Experience
XML + Semantics = DARPA Agent
Markup Language (DAML)
William Holmes, Dr. Paul Kogut
Management & Data Systems
Valley Forge, PA
June 4, 2001
JS01 June 4-6, 2001
Principals, Practice & Experience
Roadmap
 The Semantic Web
 Agents & Ontologies
 Object Management Group (OMG) Initiatives
 The DARPA Agent Markup Language (DAML)
 What is it?
 How does it fit in? / What is its role?
 LM M&DS UML-based Ontology Toolset (UBOT)
 Ontology Design & Consistency Checking
 Automated Annotation via AeroTextTM
Page 2
JS01 June 4-6, 2001
Principals, Practice & Experience
Semantic Web: The Vision
Hi Pete, it’s Lucy. I’m at the doctor’s office.
Mom needsGreat!
to seeI’lla have
specialist
and then has
my agent
to have a series
of the
physical
therapy sessions.
Hello?
set up
appointments.
Sure Lucy.
BiweeklyRING
or something.
Can ...
you split the
… RING
chauffeuring with me?
* Berners-Lee, Hendler, Lassila “The Semantic Web” Scientific American, May 2001
Page 3
JS01 June 4-6, 2001
Principals, Practice & Experience
The Vision
Lucy’s agent looks up several lists of
providers and checks for ones
in-plan for Mom’s insurance, within
Schedule a treatment plan for MomLucy’s agent formulates
schedule
aa20-mile
radius of her home, and
using Pete and my schedules. Only use
of appointments for therapists
with
with a rating of excellent or very good.
providers that are in-plan for Mom’s appointments
Lucy’s
agentthat
retrieves
information
available
fit
insurance, are within a 20-mile radius, into Pete and
about
Mom’s
prescribed treatment
Lucy’s
schedule.
and have a rating of excellent or very good.
from the doctor’s agent.
Semantic Web
* Berners-Lee, Hendler, Lassila “The Semantic Web” Scientific American, May 2001
Page 4
JS01 June 4-6, 2001
Principals, Practice & Experience
That’s Great but How?
 Need Agents
 Definition (Merriam-Webster):
 one
who is authorized to act for or in the place of another as a
business representative
 Provide a means of processing the volumes of
information found on the web.
 Need Ontologies
 Definition:
 Philosophy
- A theory about the nature of existence.
 A.I. - A formal definition of relations among terms.
 Provide a “semantic grounding” for the web.
Page 5
JS01 June 4-6, 2001
Principals, Practice & Experience
What are Agents?
 In software, “Agent” is used in many different ways:
 persistent
process/daemon:
 mobile code
 autonomous robots
 “intelligent agent” - what makes it intelligent?
 simple definitions that capture the essence of
agents:
 an
Object that decides when to say go and when to say no OMG
 “programs that operate at a high enough semantic level that they
can form new connections to other programs in order to get a job
done” Burstein, McDermott
Page 6
JS01 June 4-6, 2001
Principals, Practice & Experience
Why Agents?
 Agents are the next generation of middleware
– built on top of existing middleware (e.g., CORBA, EJB, Jini)
– run-time integration via dynamic discovery and resource negotiation
– emphasis on broker and facilitator agents (e.g. yellow pages)
 Agents are the next generation user interface
– more complex applications require personal assistant agents
– multi-modal interfaces e.g. speech, handwriting, gestures
– user specifies goals and agent handles details according to user
preferences
Internet / Intranet agents
I need to go to Fort Worth
on Monday for 3 days.
hotels
itinerary, tickets & maps
Page 7
personal assistant
agent
maps
car rental
airlines
JS01 June 4-6, 2001
Principals, Practice & Experience
Why Agents? (Cont.)
 Agents are the next level of component abstraction
 agents are components with attitudes
 beliefs,
desires, goals…*
 agents interact like humans via speech acts
 request,
inform, promise
 agents share a context for efficient communication
 domain
model ontologies are used at run-time
 ontology agent/services - query, retrieve and translate ontologies
*Labrou, Finin, Peng “Agent Communication Languages:The Current Landscape” IEEE
Intelligent Systems March/April 1999
Page 8
JS01 June 4-6, 2001
Principals, Practice & Experience
Examples of Agent Applications*
 personal assistant - digital secretary
–
travel arrangements
– meeting schedule coordination
– personalized information filtering
– mobile computing
 internet/intranet information retrieval/summarization
 electronic commerce
 enterprise workflow - e.g., sales, order processing, shipping
 military command and control
 synthetic characters (e.g., Extempo Systems, Virtual Personalities)
 robots - manufacturing, office, domestic
 design and engineering
*see Hendler “Is There An Intelligent Agent in Your Future?”
http://helix.nature.com/webmatters/agents/agents.html
Page 9
JS01 June 4-6, 2001
Principals, Practice & Experience
Ontologies
 Machine readable semantic specifications.
 Include terms, relations, and inference rules
 What does “capital” mean?
 Seat
of government (Tallahassee, Harrisburg, Austin)
 An upper-case letter
 monies, securities, investments, etc…
 the top of a column or pillar.
 XML is Not Enough!!!
 Allows definition of syntax, but not semantics (meaning)
 Can be considered the “Assembly Language” of the Web.
Page 10
JS01 June 4-6, 2001
Principals, Practice & Experience
OMG Initiatives
 OMG Agent Platform Special Interest Group (SIG)
 extend the OMG Object Management Architecture (OMA) to better
support agent technology
 identify and recommend new OMG specifications in the agent area
 recommend agent-related extensions to existing and emerging OMG
specifications
 promote standard agent modeling techniques
 see http://www.objs.com/agent/index.html
 OMG Ontology Working Group
 Align the domain modeling activities of OMG with the Semantic Web
initiative of the World Wide Web Consortium and with related
ontology development projects such as DARPA DAML and IEEE
SUO (Standard Upper Ontology).
Page 11
JS01 June 4-6, 2001
Principals, Practice & Experience
DARPA Agent Markup Language
 Machine-Readable Ontologies & Annotation (markup)
 Aimed at “Resources”, Not just web-pages
 Sensors
 Services
 Appliances
 Lots of industry Buzz*
 Scientific American
 IEEE Distributed Systems
 New York Times
 ZDNet
…
*See http://www.daml.org/inthenews.html
Page 12
JS01 June 4-6, 2001
Principals, Practice & Experience
DAML: Basic Idea
queries
DAML
web
pages
annotation
links
web crawlers
DAML
ontologies
annotate
manually or
semi-automatically
links
queries
DAML
annotation
queries
agents
Page 13
schema
RDBMS
data
web pages, databases, legacy software, devices, sensors...
have annotations linking their terms to ontologies
JS01 June 4-6, 2001
Principals, Practice & Experience
DAML Annotation: Extreme Metadata
Evolution of Metadata
explicit semantic agreements via machine-readable ontologies
implicit semantic agreements on paper!
document
parsing info
keywords
browser
web crawler
Page 14
XML
schema
Subject verb object
semantics for
selected sentences
Full semantics
for all content
XML
parsers
agents
(near-term)
agents
(future)
JS01 June 4-6, 2001
Principals, Practice & Experience
DAML Program
 Main DAML website = www.daml.org
 Duration: August 2000 to Fall 2002
 Approach:
 MIT W3C semantic web activity
 http://www.w3c.org/2001/sw/
 “The
semantic Web and its languages” in IEEE Intelligent
Systems, November/December 2000, pages 67-73 available at
http://www.ksl.Stanford.EDU/projects/DAML/
 Extend XML/RDF
 represent
ontologies
 annotate web pages and other information with links to ontologies
Page 15
JS01 June 4-6, 2001
Principals, Practice & Experience
DAML Program (Cont.)
 17 research teams and 1 integration team
 industry, academia and World Wide Web Consortium
 expertise in AI knowledge representation, logic and web technologies
 cooperation with European Union IST Program

www.daml.org/committee/
 DAML language definition
 Ontology Definition
 Rules Definition
Page 16
JS01 June 4-6, 2001
Principals, Practice & Experience
DAML Program (Cont.)
 DAML tools
 ontology development and verification
 web page annotation
 dynamic composition of agent services
 distributed query processing and inference
 ontology translation
 DAML trial applications
 Government: Intelink, Center for Army Lessons Learned
 Commercial: e-commerce, information retrieval
Page 17
JS01 June 4-6, 2001
Principals, Practice & Experience
The Origins of DAML
 Extensible Markup Language (XML)
 provides syntactic interoperability
 depends on implicit semantic agreements
 Resource Description Framework (RDF)
 designed to represent metadata for web resources in an XML syntax
 triples: <shoeGen:GovermentOrganization rdf:ID="DARPA”/>
<shoeGen:OrganizationHomePage rdf:about="http://www.darpa.mil/">
<shoeProj:authorOrg rdf:resource="#DARPA" />
</shoeGen:OrganizationHomePage>
 RDF Schema (RDFS)
 adds OO concepts: class and subclass
DAML
RDFS
RDF
* For more information see www.w3.org
Page 18
XML
JS01 June 4-6, 2001
Principals, Practice & Experience
Status of DAML
 DAML+Oil (ontology)
 released January 2001 - latest revision March 2001
 language specifications and documentation:
 http://www.daml.org/2001/03/daml+oil-index.html
 design rationale
 http://www.cs.man.ac.uk/~horrocks/Slides/index.html
 DAML-L (logic)
 rule representation and reasoning
 development in progress
Page 19
JS01 June 4-6, 2001
Principals, Practice & Experience
UML-Based Ontology Toolset (UBOT)
 We are applying:
 graphical modeling and formal verification techniques from software
engineering
 text extraction from natural language processing
 lexical semantic resources from cognitive science
 to build a tool-set that supports
 creation, extension and consistency checking of DAML ontologies
 DAML annotation of information resources for agents
 intended for users who have minimal training in knowledge
representation and agent theory
 see http://ubot.lockheedmartin.com/
Page 20
JS01 June 4-6, 2001
Principals, Practice & Experience
UBOT Team
 Lockheed Martin Management & Data Systems
 architecture, development and integration
 Versatile Information Systems (Northeastern
University)
 formal verification of UML
 Lockheed Martin Advanced Technology Center
 field test of DAML and UBOT
 Kestrel Institute
 automated formal methods
Page 21
JS01 June 4-6, 2001
Principals, Practice & Experience
UBOT Architecture: Ontology Engineering
UBOT
UML GUI
DAML Ontology
Engineer
XMI
models
XMI
models
Consistency
checking
results
UML
Formalization
Slang
models
UML
DAML
Translation
Baseline
DAML
ontologies
Extended
DAML
ontologies
Semantic
inconsistencies
Specware
Page 22
JS01 June 4-6, 2001
Principals, Practice & Experience
UBOT Architecture: Annotation
UBOT
UML GUI
XMI
UML
DAML
corrected
annotation
Translation
uncorrected
annotation
DAML
Annotator
Text or
web pages
Page 23
Extraction
to DAML
Translation
automatically
generated
Text Extraction
DAML annotated
text or web pages
DAML
Ontologies
JS01 June 4-6, 2001
Principals, Practice & Experience
UBOT Architecture: COTS Components
 UML GUI
 Tau UML Suite (Telelogic)
 Specware (Kestrel Institute)
 supports ontology consistency checking via formal methods
 SNARK theorem prover (SRI)
 Text Extraction
 AeroText (LM M&DS)
 extracts entities (e.g. people, organizations, etc.) from natural language
 recognizes relationships between entities
(e.g. [organization] hired [person] )
 developed for the U.S. Intelligence Community
 12 years experience with sophisticated linguistic processing
 many fielded applications
Page 24
JS01 June 4-6, 2001
Principals, Practice & Experience
UML GUI: Tau UML Suite
Page 25
JS01 June 4-6, 2001
Principals, Practice & Experience
Text Extraction: AeroText
Document
Window
Extraction
Display
Page 26
JS01 June 4-6, 2001
Principals, Practice & Experience
Automatic Annotation: AeroDAML
Page 27
JS01 June 4-6, 2001
Principals, Practice & Experience
Questions?
Page 28
JS01 June 4-6, 2001