1

Open Ontology Repository SessionOpen Ontology Repository SessionOOR-Team PresentationOOR-Team Presentation

Ontology Summit 2008Ontology Summit 2008Interoperability Week, NISTInteroperability Week, NIST

Gaithersburg, MD Gaithersburg, MD

MikeDean, LeoObrst, PeterYimApril 29, 2008

v 0.98

2

Agenda: Presenting the OOR Initiative

1) What is the OOR? Overview, rationale and motivations– Leo Obrst

2) What are some existing efforts? How do these address or satisfy the rationale?– Bruce Bargmeyer

3) What do users expect? How do these needs align with the rationale?– Ken Baclawski

4) How do these needs translate into OOR system requirements? How do these satisfy the rationale?– Evan Wallace

5) What is the roadmap to developing/delivering these requirements in an OOR implementation effort? How does the roadmap satisfy the rationale?– Mike Dean

3

Overview, Rationale & Motivations

Leo Obrst

4

Overview• Recognizing of the need for an Open Ontology Repository, the co-conveners got their act

together:– 2001 DAML Ontology Library (Mike Dean)– 2005 MITRE Study on OWL/RDF Registry & Repository (Leo Obrst)– 2002/2005 CIM3-CWE / CODS initiative (Peter Yim)

• 2008-01-03: Open Ontology Repository initiative - Planning Meeting– Proposed to have OOR as the theme for Ontology Summit 2008

• 2008-01-23: OOR Initiative - Founding Members Conference Call– “Open Ontology Repository (OOR) Initiative” came into being, with about 40 participants (active

participants, as well as observers)

– Team adopts Mission Statement• 2008-02-07: OOR team adopts their “Ontology Repository” definition• 2008-02-28~04.10: joined with the OntologySummit2008 effort and co-organized four OOR-

Panel Sessions: – Covering: “Technology Landscape,” “Expectations & Requirements,” and “Ontology of Ontologies”

• OOR team virtual activities being hosted within the Ontolog collaborative work environment, for the time being

5

The charter of the Open Ontology Repository (OOR) Initiative is to the promote the global use and sharing of ontologies by:

• 1. establishing a hosted registry-repository;• 2. enabling and facilitating open, federated, collaborative

ontology repositories, and• 3. establishing best practices for expressing interoperable

ontology and taxonomy work in registry-repositories.where, “An ontology repository is a facility where ontologies and

related information artifacts can be stored, retrieved and managed.” -- definition as adopted by the OOR-team / 2008.02.07

Homepage: http://ontolog.cim3.net/cgi-bin/wiki.pl?OpenOntologyRepository

6

Rationale & Motivations (1)

• Why are we interested in an OOR and what purpose does it serve?• Isn’t the Semantic Web notion of distributed islands of semantics sufficient as

a de facto repository?– If you put it out there, will they come?– If you build it better and put it out there, will they prefer yours?– History does not show this laissez faire “field of dreams” is good reality– The "clickable" web has been very successful in employing this strategy for html

documents– However the use and content of the semantic web has different characteristics that make

it far less tolerant of the change and frequent errors which are commonplace on the clickable web.

• Distinguishing characteristics of the Semantic Web– Machines rather than humans are the primary consumers of content. Errors that a human

may be able to diagnose and fix (such as a change in location of a document) are likely fatal for machine processing

– The use of owl:imports creates a strong transitive dependency between ontology documents; changes in any imported document (imported directly or through nested import) can cause the resulting import closure to be inconsistent or to change its meaning or computational characteristics significantly.

– Ontologies convey a precise meaning with an unambiguous machine interpretation. This means that, when using this content, careful selection and precise reference is critical.

7

Rationale & Motivations (2)

• Value Added to the content by an Open Ontology Repository/Registry:– The OOR is reliably available– The OOR is persistent and sustainable, so you can be confident when committing to its use– The OOR has information about when, why, and how an ontology has changed, so you can

be aware of changes that may effect its usability– You can find ontologies easily– Ontologies are registered, so you know who built them– Metadata provides the ontology purpose, KR language, user group, content subject area, etc.– The OOR includes mappings, so you can connect ontologies to other ontologies– The OOR content has quality and value, as gauged by recognized criteria– The OOR supports services, so that ontologies can map and be mapped, find and be found,

can review/certify and be reviewed/certified, can hook your own services into and can use the services others have hooked in

– Ontologies can reuse or extend other ontologies, including common middle and upper ontologies

– The OOR can be easily extended• Ref. also – opening post to the OOR-team ref “definition: registry vs. repository; goals, etc.” -

http://ontolog.cim3.net/forum/oor-forum/2008-01/msg00016.html

8

Open Ontology Repository

User Needs&

Requirements

Ken Baclawski

9

“OOR” - what is in scope?• Repository: "An ontology repository is a facility where ontologies

and related information artifacts can be stored, retrieved and managed"

– first: the persistent store for ontologies– then, the registry for ontologies in the repository– then progressively, the value-added services

• Ontology:– all types of artifacts on the ontology spectrum

• from folksonomies, terminologies, controlled vocabularies, taxonomies, thesauri, ... to data-schema, data-models ... to OWL ontologies ... and, axiomatized logical theories

• from shared understanding ... to ontological commitments ... to the future of standards

• Open: – open access; compliance with open standards; open technology (with

open source); open knowledge (open content); open collaboration (with transparent community process)

– open to integration with “non-open” repositories via an open interface

10

OOR Users Needs (1)

• Who are the users of an OOR– ontology developers (individuals or distributed teams)

– ontology centers and institutions

– end-users (human) who need to search/browse an ontology

– software agents (machine) who need to use the ontologies

– application developers

• When– design time

– run-time (dynamic, real-time, on-the-fly, ...)

11

OOR Users Needs (2)

Through the two virtual panel sessions and our online discourse, we heard from experts among the panelists and participants coming from different domains:

• 2008_03_27 - Thursday: Joint OOR-OntologySummit2008 Panel Discussion: "An Open Ontology Repository: Rationale, Expectations & Requirements - Session-1" - Chair: LeoObrst & FabianNeuhaus; Panelists: WilliamBug, EvanWallace, JohnLMcCarthy, KenBaclawski, PeterBenson & RexBrooks - http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2008_03_27

• 2008_04_03 - Thursday: Joint OOR-OntologySummit2008 Panel Discussion: "An Open Ontology Repository: Rationale, Expectations & Requirements - Session-2" - Chair: LeoObrst & FabianNeuhaus; Panelists: DougLenat, DekeSmith, MarciaZeng, DeniseBedford, PatHayes, MalaMehrotra & RobRaskin - http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2008_04_03

12

OOR Users Needs (3)

The top needs came out to:• that there is a well-maintained persistent store (with high availability and

performance) where ontological work can be stored, shared and accessed• having “ontologies” properly registered and “governed,” with provenance and

versioning support, and made available (logically) in one place so that they can be browsed, discovered, queried, analysed, validated and reused

• allow ontologies to be “open” and unencumbered by IPR constraints, in terms of access and reuse

• services that can be provided across disparate ontological artifacts to support cross-domain interoperability, mapping, application and making inferences … and having such semantic services be properly registered and available to support peer OORs

• (in addition to the panel proceedings cited above) ref. IM chat/discussion: http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2008_03_27#nid1CJJ and, for example, input from AndrewSchain (NASA/HQ): http://ontolog.cim3.net/forum/ontology-summit/2008-04/msg00010.html

13

A sample of the input on Needs and Expectations

… based on summary slides received from some of the OOR-Panelists on the 2008.03.27 &

2008.04.03 “Requirements Panel” Sessions

14

What We’d Want a Good Host to Provide• A commitment to use – to have contributors all provide content under – some Creative

Commons license, as opposed to e.g. a GNU license

• Retention of the provenance/lineage of contributed ontological content

• Agreement on some of the most fundamental ontological relations

• Agreement on a small set of inter-ontology alignment relations

Doug Lenat

CycorpDoug@cyc.com

Content that Cycorp could provide to be be hosted: * OpenCyc (www.opencyc.org) 100% free even for commercial purposes * ResearchCyc (researchcyc.cyc.com) free for R&D purposesIn both cases, there are ontologies plus inference engines and API-level and graphical interface tools

Meta-level message: Look at OKKAM, LarKC, etc., and decide what role, if any, OOR can/should play, and how it should tie in with those other efforts.

15

What’s in OpenCycDoug Lenat

• (#$isa 596215)

• (#$genls 99198)

• (#$disjointWith 6114)

• (#$resultIsa 4277)

• (#$resultGenl 1206)

• (#$argIsa 35617

• (#$argGenl 5398)

• (#$arg1Isa 16748)

• (#$arg1Genl 2354)

• (#$arg2Isa 14114

• (#$arg2Genl 2283)

• (#$arg3Isa 3486)

(#$argFormat 5493)

(#$arg2Format 3320)

(#$functionalInArgs 1427)

(#$arity 16416)

(#$arityMin 958)

(#$comment 57305)

(#$genlPreds 7440)

(#$negationInverse 990)

(#$genlMt 26078)

(#$denotationInEnglish 409745)

(#$synonymousExternalConcept 13916)

Explicitly: 300k terms; 14k predicates; 57k classes; 2 million assertionsImplicitly: There are infinitely more nonatomic terms and inferred assertionsMore subtle but crucial point: There are infinitely many contexts (microtheories) defined compositionally rather than having only explicitly reified contexts

This means there are 596k “isa” assertions in OpenCyc

E.g., mapping between a term in

OpenCyc and a WordNet synset

“What is impossible to do right now, but, if you could do it, would fundamentally change your business?”

1990 Joel Arthur Barker

Codification at source !– Common metadata (ISO 22745-20/eOTD)

• an end to data mapping

– Requirement specifications (ISO 22745-30/eOTD-i-xml )

• an end to incomplete data

– Data provenance (ISO 8000-120)• an end to inaccurate information

Vision of the Future Peter Benson

NATO codification system as the foundation for the eOTD, ISO 22745 and ISO 8000

Faster Faster datadata – Better – Better datadata – Cheaper – Cheaper datadata

Codification at Source - Peter BensonUsing standards to automate the data supply chain

Data requestor

Data provider

Sub

eOTD-i-xml(data requirements statement) ISO 22745-30

eOTD-q-xml(query)

ISO 22745-35

Sub-TiereOTD-q-xml

Sub-TiereOTD-r-xml

eOTD-r-xml(data exchange)ISO 22745-40

Faster Faster datadata – Better – Better datadata – Cheaper – Cheaper datadata

18

Rex Brooks: Content Provider-Repository BuilderFocus on Architecture, Registry-Repository & Emergency Data Exchange Language Reference Information Model

(EDXL-RIM)

19

Rex Brooks: Content Provider-Repository BuilderFocus on Architecture, Registry-Repository & Emergency Data Exchange Language Reference Information Model

(EDXL-RIM)

20

Developer Requirements - Ken Baclawski

Must have the ability to browse and query small segments of an ontology.

Good to have the ability to dynamically curate and suggest changes via the user community.

Ideally, it can be used to navigate across inferred information that is associated with a small set of terms and that comes from many ontologies.

21

End User Requirements - Ken Baclawski

Must have– Ability to efficiently navigate multiple hierarchies– Consistency across multiple ontologies

Good to have– Ability to provide live feedback– Allow annotating relationships or propose new terms

Ideally, it can– Support scientific hypothesis testing

22

OOR needs for content /application providersMala Mehrotra

• Content developers: Discover related terms/axioms/models for reuse– Context – collaboration groups of concepts

• region (geographic, biological, political)

– Depth/detail • month in SUMO vs. monthDescription in DAML time ontologies

– Differences in competing models• TimeInterval in SUMO vs DurationDescription in DAML

– Degree of Crossover/Overlap • More than just imports closure

• Orthogonality measures across ontologies

• Application developers: Interoperate using multiple ontologies– Create formalized mapping relationships

– Find mapping relationships

23

Infrastructure Needs Mala Mehrotra

• Cognitive Tools for discovery

– Collaborating groups of concepts used in applications

– Implicit relationships across resources

– Ontological/Taxonomy hierarchy browsing

– Human-machine collaboration mode

• Mapping Tools for capturing inter-resources’ relationships

• Need formal representation of relationships for reasoners

– A large repertoire of relationships

– Multiple ontological representations

– Mechanisms to represent formalism in human-readable form

24

Needs vs. Rationale• The “Needs and Expectations” map well to the “Rationale”

cited in the previous slide• A community OOR will provide us with

(from slide #11 from Denise Bedford's 2008.04.03 brief)

– Knowledge value– Collaboration value– Shared process value

• However, further to the intellectual discourse on what an OOR should be, the implementors of the OOR will also need to answer questions like:

– How could we make sure the OOR is still around in 100 years?– What can be done about assuring the sustainability of the resources,

expertise, quality of the ontologies in the OOR and the services provided?– How can we ensure long term value, commitment and continuous

improvement to the OOR?

25

Open Ontology Repository

Translating User Needs intoRequirements for

OOR Implementation

Evan Wallace

26

Translating Needs to Requirements• Active discussion on the matter evolved on the

[oor-forum] list, initiated from threads like:– post from EvanWallace (NIST):

http://ontolog.cim3.net/forum/oor-forum/2008-04/msg00011.html .. & – post from ToddSchneider (Raytheon):

http://ontolog.cim3.net/forum/oor-forum/2008-04/msg00012.html

• breaking it down to:– general requirements (scalability, distributed repository

support, platform independence, ...)– requirements to support search and discovery– requirements to support subscription and notification– management requirements, &– governance requirements

27

Requirements

• Evan to possibly add next level of “system requirements” details here (culling from the list discussion) - optional

28

To Enable the Management and Services

specified in the Requirements

we need to capture a set of metadata about the ontologies

… here are some of the input from the 2008.04.10 Panel Session

On “Ontology of Ontologies”

29

Objective Michael Gruninger

• This is based on the communique from the 2007 Ontology Summit that took place April 22-23, 2007 in Gaithersburg, Maryland, at the National Institute of Standards and Technology (NIST).

• Provide a framework that ensures that we can support diversity without divergence, so that we can maintain sharability and reusability among the different approaches to ontologies.

• To this end, we can define a set of characteristics common to all approaches and then propose a set of features that can be used to distinguish among different approaches.

30

DimensionsMichael Gruninger

• We can identify a set of dimensions that can be used to distinguish among different approaches to ontologies.

• There are two kinds of dimensions:• Semantic - how an ontology specifies the meaning of

its vocabulary– Expressiveness of the ontology representation language– Level of structure– Representational granularity

• Pragmatic - the purpose and context in which the ontology is designed and used

– Intended use– Role of automated reasoning– Descriptive vs prescriptive– Design methodology– Governance

31

OMV - Ontology Metadata VocabularyPeter Haase

• OMV is … a metadata schema– Captures reuse-relevant information about an ontology

• OMV consists of … core and extensions– OMV Core: fundamental information about an ontology and

its life cycle– OMV Extensions: detailed account on specific phases of an

ontology life cycle

• OMV is designed … as an ontology• OMV is realized … in OWL DL

• Website http://omv.ontoware.org/

32

Applications of OMV Peter Haase

• Numerous existing and planned applications of OMV– Ontology Registries in NeOn

• Oyster as Open Source implementation• Centrasite as commercial product of Software AG

– Watson - Gateway to the Semantic Web• Web interface for searching ontologies and semantic documents

– Stanford BMIR intend to use OMV in Protege and their Bioportal ontology repository

– OMG intend to use OMV in their ontology repository

• OMV development sustained by OMV Consortium– Current members: UPM, AIFB, TU Berlin, Stanford BMIR– Looking for wider adoption / standardization in the community– Opportunity to join, contribute, collaborate!

33

Requirements vs. Rationale• Again, the “Requirements” map well to the “Rationale”

cited in the previous slide, additionally, though

• Unlike R&D, what's not “interesting” there could still be crucial in implementation and in production

– for example delivering high availability, performance, even security, spam control ... these are almost irrelevant to “serving ontologies,” yet essential to having a viable OOR

• For the *real* implementation of the OOR initiative, we would still need:

– the proper Organizational Model that would deliver the previously mentioned needs (like sustainability, ... etc.)

– a viable Operating (Business) Model, and– most importantly, the community, and the skills and generosity of its

membership, to get this off the ground

34

Existing Efforts

Bruce Bargmeyer

… based on input from the 2008.02.28 OOR-Panelon “Ontology Registry and Repository Technology

& Infrastructure Landscape”

35

eXtended Metadata Registry (XMDR)

Bruce BargmeyerWhat XMDR Brings to the Table:• Use cases - semantics challenges - and Requirements• Potential design specifications

– Proposed specifications for ISO/IEC 11179 Edition 3 – A UML Model, definitions, and OWL ontology

• Modular software architecture and open source software modules• Open Source XMDR software• Test content – concept systems including thesauri, taxonomies,

ontologies• A group of participants (XMDR project) that has considerable

experience in this area. See: XMDR.org

36

Modular XMDR ArchtitectureBruce Bargmeyer

Registry Store

Search & Content Serving (Jena, Lucene)

XMDR metamodel (OWL & xml schema)

standard XMDR filesstandard XMDR files

standard XMDR filesstandard XMDR files

LogicIndex

Content Loading & Transformation

(Lexgrid & custom)

Human User Interface(HTML fromJSP and javascript; Exhibit)

Metadata Sources concept systems,

data elements

USERSWeb Browsers…..Client

Software

Application Program Interface (REST)

Authentication ServiceValidation

(XML Schema)

MappingEngine

Logic Indexer(Jana & Pellet)

Text Indexer(Lucene)

Metamodel specs(UML & Editing)

(Poseidon, Protege)

XMDR data model & exchange format

XML, RDF, OWL

TextIndex

Postgres Database

Third Party Software

37

DAML Ontology LibraryMike Dean

• Created early in the DARPA Agent Markup Language program– Organize content– Promote reuse– Demonstrate adoption

• 282 DAML+OIL and OWL ontologies submitted from October 2000 – December 2003

• Cited in several ISWC papers– Property (Feature) use across libraries

• Largely replaced by Ontaria and SchemaWeb• Available at http://www.daml.org/ontologies/

– daml.org now archived at W3C

38

Lessons Learned Mike Dean

• Curation/quality control– Many users desired some indication of use and

quality of the ontologies, e.g. user ratings

• Cacheing– Many links subsequently became unavailable– Desirable to store (and make available) local

copies

39

NCBO BioPortalMark Musen

• The National Center for Biomedical Ontology (http://bioontology.org) is developing BioPortal, an open-source repository of ontologies, terminologies, and thesauri of importance in biomedicine.

• An early version of BioPortal is accessible at http://bioportal.bioontology.org. Users can access the BioPortal content interactively via Web browsers or programmatically via Web services.

40

BioPortal will offerMark Musen

• Ontology repository functionality• Linkages among different ontologies• Community-based peer review and ontology

annotation• Linkages between ontology content and

related online data repositories• Support for communities of ontology users

and developers

41

OASIS ebXML RegRepFarrukh Najmi

• A generic registry / repository standard

– ebRIM = meta infomodel, ebRS = services and protocols

– Approved as OASIS and ISO standards

– Version 4 of RegRep expected in 2008

– Highly extensible

• Not specifically an Ontology repository, but can be made so

– ebXML RegRep Profile for OWL-Lite is an approved specification

• Has rich feature set to support use cases and architecture

– Extensible metamodel, extensible protocol, stored queries, extensible relationships, service model, validation, cataloging, subscription & notification, role-based access control and authorization, change history, federation / federated query, SOAP and REST bindings, Java API (JAXR)

• freebXML Registry provides a royalty-free open source implementation

• Is more a toolkit than an out-of-box solution

42

ebXML RegRep as an OOR Server: A Proposal - Farrukh Najmi

• Build upon RegRep 4.0 impl from Wellfleet Software

• Implement OWL-Lite Profile (modulo RegRep 4)

• RegRep does not provide Ontology specific UI, use Protege

• Integrate RegRep 4.0 with Protege such that

– RegRep serves as backend for Multi-user Protege client

– Protege reasoning engine serves as Reasoning plugin for RegRep

• Initially deploy a single Root OOR instance with pilot users playing various roles in the collaborative ontology management use cases

– Use OpenID as distributed identity management solution

• Later facilitate deployment of Community-specific OORs (e.g. Medical, GIS, Defense ...)

43

What Does CIM3 Do?Peter Yim

• Mission: to enable more effective distributed collaboration and virtual enterprise through bootstrapping collective intelligence over the Internet

• Doing business as: “cim3.com”, “cim3.net” and “cim3.org” cim3.com – the business arm of the company cim3.net – the collaborative work environments where client

Communities of Practice and distributed team workspaces are hosted cim3.org – the research arm, and holder of the company’s open

technology, content and other intellectual properties

• Products/Services: providing an ISP/ASP based Collaborative Work Environment (“CWE”) infrastructure that enables distributed project teams, virtual enterprise partners and communities of practice to work effectively over the Internet.

44

CIM3's potential role in OOR - PeterYim

• provide the “plumbing” (the bottom layer of the technology stack) - a robust hosted (hardware and network) infrastructure for OOR– Network Facility:

• Tier-1 IPv4 Internet hosting facility (IPv6 ready)• 100Mbps bandwidth into the Internet backbone

(upgradable to 1Gbps in short order)

• Backbone: multiple OC192 & Gige self-healing fiber-ring (among the top 10 networks in the world as measured by connectivity to the rest of the Internet.)

– Linux Servers (mainly on IBM 1u boxes)

• Triple redundant storage (in 2 locations)• Locked-down system environment, with spam-filtering and

content filtering capabilities

• provide a collaborative work environment for the OOR team• help facilitate the distributed teamwork - project coordination

and management

45

Open Ontology Repository Roadmap

Mike Dean

46

OOR Is …

• An open source software platform

• 1 or more public instantiations of that platform

• A sustainable organization

• (Lots of potential parallelism here)

47

Apache-like Software Platform

• Architectural framework (internal APIs, core representation standards, processing pipeline)

• A few core modules (basic registry, GUI, web service interfaces, …)

• Lots of optional modules (pick and choose when instantiating)– Quality and gatekeeping (basic checks, usage-based, community

ranking, curation, etc.)– Languages (OWL, RDFS, Common Logic, UML, SKOS, etc.)– Mapping and translation– Federation (bi-directional, one way)– Repository (expanded persistence)– Editing (access control, versioning)– Encapsulations of existing ontology services– …

48

OOR Federation

• Other OOR instances– P2P (ish)– Easiest case - we have full control

• Collaborative ontology editing environments (knoodl, Semantic MediaWiki, BioPortal, CODS, etc.)– Want to cooperate rather than compete

• Other registries/repositories• “Loose” ontologies posted on the WWW

– Add metadata and apply services

49

“Open”

• OOR platform software should be open source (like Apache)– Probably use one of the licenses from opensource.org

• OOR instantiations can set their own policies with respect to ontology content licensing

• Accommodates– Open source content– Private instantiations behind firewalls– Commercially licensed content (e.g. ResearchCyc) and

services• OOR organization should do whatever it can to promote

the adoption of ontologies and related technologies that benefit the community

50

Public Services• At least one “central” instantiation

– Showcase for the technology• Employ many/all optional modules

– Should include some sort of “registry of OORs” or OOR DNS

– Might federate ontologies from domain-specific public OORs– Primary focus on “open source” ontologies – <OOR.org> is already taken; Mike Dean has registered

OpenOntologyRepository.org & OpenOntologyRepository.net– Peter Yim has volunteered supporting infrastructure, initially

51

OOR Organization Support

• Volunteers (now)• Funded OOR proposals (hopefully soon)• Collaborative projects/proposals citing OOR

– Enhances credibility of both parties– Most immediate path forward

• Endowment• Acquisition by another sustainable organization

(Apache Foundation, W3C, OMG, …)• Should adopt a W3C-like persistence policy for

software and content

52

Linking Open Data• Ontologies as (meta) data

– Adopt their conventions• Nice visual depiction of

increasing inter-connections

• A community linked by best practices and a cool logo

53

Strawman Phase 1

• 6 month time horizon• Basic registry and repository

– Might build off XMDR• Some federation capability• Initial versions of architectural framework and

core modules• Support for at least 1 language

– Other optional modules as they become available• Specific open source license selected• 1 or more instantiations

54

Going Forward• oor-forum@ontolog.cim3.net email list

– http://ontolog.cim3.net/forum/oor-forum/ archives

• Next OOR telecon– Friday, May 9, noon EDT

• Calling for:– Collaborators / contributors– Developers– “A cool logo!”

55

Open Discussion

Q & A