1

DARPA Agent Markup Language(DAML)

Dr. Mark Greaves

May 2005

Ontologies for the Web

Jun 2000 Program start

Feb 2004 OWL accepted by W3C as a Web Standard

Dec 2004 SWRL FOL and OWL/S submitted to W3C

May 2005 Program complete

2

What is DARPA?

DARPA = Defense Advanced Research Projects Agency

Long Range R&D Organization of the US Department of DefenseEstablished 1958 as a US response to the Soviet launch of SputnikPursues high-risk, high-payoff basic and applied researchOrganizationally part of USD(AT&L) and DDR&EOperates in coordination with, but independent of, the military research and

development establishment (ARL, AFRL, ONR)Committed to maintaining U.S. military technology superiority

Chartered to Prevent Technological SurpriseFunds work that is a counterpoint to traditional thinking and approachesNoteworthy programs include VELA HOTEL, M-16, Stealth aircraft, GPS,

ARPANET, Unmanned aircraft, most early AI, MEMS, ARPANET…

FY05 research budget is ~3B

DARPA Description and active solicitations at www.darpa.mil

3

DARPA, DAML, and Google…

#2

#3

Google “darpa”on 10/21/04

4

DAML Program Summary

Solution:Augment the web to link machine-readable knowledge to web pages

– Extend RDF with Description Logic

– Use a frame-based language design

– Create the first fully distributed web-scale knowledge base out of networks of hyperlinked facts and data

Approach:Design a family of new web languages

– Basic knowledge representation (OWL)

– Reasoning (SWRL, OWL/P, OWL/T)

– Process representation (OWL/S)

Build definition and markup tools

Link new knowledge to existing web page elements

Test design approach in the Intelligence Community

Standardize the new web languages in the W3C

People use implicit knowledge to reason with web pages

People use implicit knowledge to reason with web pages

Computers require explicit knowledge to reason with web pages

Computers require explicit knowledge to reason with web pages

Existing Web

(HTML/XML over HTTP)

Semantic Web

(OWL over HTTP)

Links via URLs

Problem: Computers cannot process most of the information stored on web pages

5

Program Elements

• Web Ontology Language (OWL)– Enables knowledge representation and

tractable inference across the web

– Based on Description Logics and RDF

• OWL Reasoning Languages– SWRL Rules Language: Supports business

rules, policies, and linking between distinct OWL ontologies

– OWL/P Proof Language: Allows software components to exchange chains of reasoning

– OWL/T Trust Language: Represents confidence that OWL and SWRL inferences are valid

– Based on Description Logic Programming

• Semantic Web Services (OWL/S)– Allows discovery, matching, and execution of

web services based on action descriptions

– Unifies semantic data models (OWL) with process models (Agent) and shows how to dynamically compose web services

– Based on process algebra and NIST PSL

• OWL Tools

Completed standards process

Started standards process

Under development

SWRL: RulesOWL/P: Proof

SWRL: RulesOWL/P: Proof

OWL/S:Semantic Web

Services

OWL/S:Semantic Web

Services

Web OntologyLanguage (OWL)

Web OntologyLanguage (OWL)

OWL/T:Trust

OWL/T:Trust

DAML Program Technical Flow

Each DAML Program Element includesspecifications, software tools,

coordination teams, and use cases

6

2004 Technical Progress

• Web Ontology Language (OWL)– W3C accepted OWL; formed Semantic

Web Best Practices Working Group to maintain the standard

– W3C agreed to host Ontaria, a permanent public OWL ontology registry/download site

– OWL gained traction (250K RDF/OWL pages, 20M+ triples, 10K classes available on-line)

– W3C Workshop on OWL in life science

• OWL Reasoning Languages– SWRL 0.6 released 24 May 2004 by the

US/EU Joint Committee; being tested at JWAC, IMO, NSA

– SWRL-FOL submitted to the W3C

– SweetRules complete

– W3 Rules Workshop April 27-28

• Semantic Web Services– OWL/S Web Services Specification

submitted to W3C

– Semantic Web Services Interest Group chartered by W3C

– Semantic Web Services Initiative (~45 organizations) coordinates commercial, DAML, and EU Framework 6 output

– SWSL and SWSL-FOL submitted

– W3C SWS Workshop June 9-10

• OWL Tools– DAML sponsored a new open source

website www.semwebcentral.org

– Over 70 OWL tools released by DAML contractors

– New OWL plugins for Eclipse

– Currently 84 hosted projects, 3M hits and >100GB of downloads since Dec 2003

7

Transition

Intelligence Community[6 funded pilots at different IC agencies]

DoD AF AMC Foreign Clearance Guide FCS SOSCOE OWL/S use in the TINAF AMC NOTAMs Joint Explosive Ordinance Detection ACTDDISA Discovery Metadata Repository Center for Army Lessons Learned Prototype

FederalCIO Council Semantic Interoperability Community of Practice formed, 2 conferences SWANS conference April 7-8 2004 (300+ attendees, 40 trade show participants)

Commercial43 companies in SWSI working on OWL/S 19 commercial OWL implementations including IBM and HP

More evidence of uptake…58 “Semantic Web” books on Amazon.comNCI Thesaurus is 100% OWLNIH and NIST are sponsoring work to define a comprehensive protein chemistry taxonomyDARPA XG using OWL for policy language vocabulary

8

DAML Schedule

• Complete OWL versioning tools, Ontaria, OWL/T 1.0• Deliver OWL/S 1.1 to W3C and complete OWL/S editors • Complete SWRL 0.6 reasoning environment and submit SWRL FOL• Tools and Outreach

– Semantic Web Applications for National Security (SWANS) and SWIG meetings– Stabilize and transfer semwebcentral.org and daml.org to W3C– Complete SWeDE, IE plugin, and reference application

Meetings and Reviews

Web Ontology Language

Trust Specification

Program Elements

PI MeetingPI Meeting

SWMU SWMU

PI MeetingPI Meeting

SWMU

Revisions (DAML+OIL), OWL Lite,OWL DL, OWL Full

FY02 FY03 FY04 FY05

PI Meeting

Semantic Web Services Semantic WebServices Initiative

Process Representation, Brokering, Profile, Grounding Ontologies

Trust Algorithms

Proof Specification

Rules Specification Createv 0.6

Logic Mappings, Descriptive Logic Programming, Tool Development

Proof Language and Query Engines

IC Transitions

SWANS Mtg

PI Meeting

W3CDelivery

FY01Work

Createv 0.7

ExternalConf

END

Ontaria,OWL Versioning

Program Milestone

FY05 Remaining Tasks

Saturn CombineHorus

NOTAMS NGA

Saturn II

JWAC ALVJWAC

SWRLReasoners

OWL/SEditors

9

DAML’s Legacy

Success = creating the conditions for early adopters to allow the semantic web revolution to succeed

DAML has had incredible successWe have gone from DARPA-hard challenge to accepted industrial standard in four

yearsThe PM has lost control of the technology

It is time for OWL to leave the DARPA nest and flyThere is more work to be done: OWL 2.0, Semantic Web Services, Rules, Query

Languages, Tools, Documentation, Killer Apps, Proof Exchange, TrustDomain-specific ontologies and applicationsMore standards, collaboration with Europe, funding organizationsMore nonacademic conferences

DAML’s intellectual thread will be carried by other programs and organizations

So… What kind of new DARPA program would compliment DAML?

10

How Does a New DARPA Program Start?

New Programs Must result in or point to a new military capability

Must be about removing a technological barrier, not a policy barrierProblem must be “DARPA-hard”; typically 10x improvement

Barrier to capability must be primarily technical, not policy

Must start from a specific new immature technology idea or ideasSpecific = must be identified at the program approval phase

New = typically based on work that is < 5 yrs old

TechOffices

SystemsOffices

ProgramOffices

6.4Prototype

6.1Tech IdeaDARPA

OLD NEW

“The Particle Accelerator”Technology Base

Syst

ems

Tech

nolo

gy

Syst

ems

Tech

nolo

gy

Syst

ems

Tech

nolo

gy

…

11

The Heilmeyer Catechism

What are you trying to do? Articulate your objectives using absolutely no jargonExample: “take anthrax off the table as a threat to our forces”

What is the new military capability that your technology could provide?

How is it done today, and what are the limits of current practice?Why is this specifically a technology problem?

What's new in your approach and why do you think it will be successful?All software is Turing-equivalent, so software methodology is usually not relevant

What is your argument/analysis that a 10x difference in a technology will result in a new capability?

Who cares? If you are successful, what difference will it make?Who is the customer for the new idea, and what evidence do you have that any

transition will be successful?

What are the risks and the payoffs? How much will it cost? How long will it take?

What are the midterm and final exams to check for success?Metrics and experimentation plans must be defined up front

12

Other Program Questions

What is DARPA’s Transition Strategy?How does new capability transfer to a Service or Agency?

Gold: DARPA work leads to a direct acquisition

Silver: DARPA work leads to a direct maturation effort by a DoD PEO

Bronze: DARPA work leads to a new capability that a contractor will try to sell back to DoD

Tin: DARPA work leads to a better state of the world

Is there an MOU / MOA and funding in the POM?

Why is this different from other DARPA and DoD programs?

What are our metrics for measuring our progress?Always difficult for software; exceptionally difficult for architectures

What are the phases of the Program?Phase I is typically 12-18 months

Phase II funding is contingent on meeting specific agreed-upon phase I milestones

13

Program Creation Basics

DARPA PM finds new technology idea(s) and links it to capability

Seedling funding to explore idea and create program briefTypically $200K - $300K / 4-6 months / 1-3 contractorsSolidify program argument, financials, milestones, phases, metrics, experimentation

strategy, and program deliverable/transition/MOUsSeedling output is the newstart brief – not jumpstart technology

Brief to DARPA DirectorRepeat a few times

Solicitation construction and publication

Source Selection (and possible plan revision)Multiple contractors, teams, areas of expertise

Contracts Awarded via an AgentProgram Phase I with milestonesDARPA Director Brief for go/no-goProgram Phase II with milestones

14

Sample Program: Dynamic User Interfaces

MAIN OBJECTIVE

EXPECTED IMPACTTECHNICAL APPROACHUse situation theory to quantify the information content of a UI

Decompose user’s info tasks into UI task specifications

Leverage OWL to create a tractable logic language that can express analytic tasks and data semantics

Apply constraint-based solvers, CBRs, and other planning technologies to yield task-specific UI specs

Map UI tasks onto available graphical elements

Build a semantically characterized set of UI graphical elements by using OWL/S and SWRL

Use a planner/shape grammar and machine learning to derive the UI layout for an user’s individual profile

Dynamically create the new UI on the user’s desk

Replace current mass-produced general-purpose UIs with task-sensitive, user-specific interfaces

Customize each user’s I/O with the data sources

User interacts with the web at the problem level

User does not have to master all the data sources and algorithms that are available

UI is automatically built for each user’s unique cognitive/perceptual talents, training, experience, and current problem context

Allow UIs to better support independent hypothesis generation and unconventional concept exploration

Faster and higher-quality analytic output

Embrace individual styles and competencies

Tune core UI planners to allow rapid confirmation or disconfirmation of different uncommon hypotheses

Increase user satisfaction

Increased agility in response to new missions

Restructure planners and interfaces on the fly to handle new information requirements and data sources

Interaction with the DBs structured around user task requirements, not data structures

Late binding the UIs relative to the individual user, task, and problem context allows for rapid learning and evolution of interface paradigms

NGA Task SME

Cognitive Task Analyst

Interface NGA Analyst

Interface Designer

Computable Interface Spec

Interface Generator

Machine learningInterface Generator reasonsover domain and task semantics to dynamically adapt interface to the situation

Interface is automatically tailored for each user

Interface adapts to the user’s context

15

How Is DARPA Different?

Lightweight and nimble organizational model“120 PMs with a common travel agent”Currently organized into 8 tech offices plus the Director

Technology – DSO, MTO, IPTO; Systems – ATO, TTO, IXO, SPO, J-UCASOffices come and go fairly frequently and the tech/systems boundary is fluidNo institutional incentives to collaborateNo technical interdependencies

No dedicated facilities beyond simple office space in Arlington, VAhttp://www.darpa.mil has programs, solicitations, lists, areas of interest

4-year personnel rotation policy embedded in the cultureNo institutional biasesNo empire building

Always looking for new Program Managers with great technical ideasPMs come from academia, industry, government, militaryMust be a US citizen with the ability to hold a clearanceMust be willing to work incredibly hard, travel extensively, and have a national-scale

visionYou will be changed by the experience, and you might change the world

Come Join Us!