Part III Putting Ontologies to Work

1
Part IIIPutting Ontologies to Work

• Examples of use by us.
• Ontologies and the Grid.
• Survey of other examples.
• The Story of Gene Ontology.

2
What are they used for taster

• A community reference.
• Support for community practice.
• Specification of database or database content.
• Controlled vocabulary for annotation.
• Application database interoperability.
• Consensual vocabulary.
• Searching, query formation indexing.
• Classification, interface/portal driving.
• Agent or service metadata management.
• Grid services description, advert, discover.

3
What are they used for taster

• Support for knowledge intensive applications.
• Text extraction, decision support, resource
  planning, intelligent interfaces.
• Knowledge repository structure.
• Knowledge acquisition.

4
Examples of ontologies in use by us

• AKT reference ontology for Depts of CS
• http//www.aktors.org
• COHSE Open Hypermedia linking using ontology
• http//cohse.semanticweb.org
• Geodise e-Science design optimisation ontology
• http//www.geodise.org
• Web services information integration study
  Qinetiq
• myGrid e-Science Grid Services ontology.
• http//www.mygrid.org.uk
• TAMBIS repository integration using an ontology
• http//www.cs.man.ac.uk/tambis/

5
The Semantic Web
OILEd, Protégé
OntoPortal, SEAL, COHSE
DAMLOIL
FaCT, Racer
RDF
COHSE
CREAM
http//www.semanticweb.org
6
Agents on the Web
Studer2000
7
The AKT project - Knowledge Services on the
Semantic Web

• Content with extensive markup and annotation
• Markup will be significantly driven by ontologies
• Services will exploit this markup to deliver
  knowledge - the right content, to the right
  person/system, in the right form at the right
  time
• Navigating
• Semantic querying
• deliver integrated answers extended by derived
  facts

8
Ontologies The Semantic Backbone
9
Ontologies Putting them to Work
10
Communities of Practice

• Objective
• Increase skills
• Agents
• Homogeneous
• Cognitive ability
• Accumulate circulate best practice
• Recruitment
• Self selecting
• Knowledge production
• Learning through the work
• Community glue
• Common passion

11
The Community of Practice (COP) Example
12
Visualising COPs
13
Integrating Knowledge Services through ontologies
14
COHSE demo http//cohse.semanticweb.org

• Annotation of web resources with DAMLOIL
  concepts from ontologies
• Dynamic linking of resources based on concepts
  and reasoning.

Linking
Annotation
15
COHSE
16
Web Services
17
Semantic Networking
The interoperability layer migrates from the
syntactic to the semantic
18
Service/Resource Descriptions

• Automated gt
• Discovery Search
• Selection
• Matching
• Composition
• Interoperation
• Invocation
• Execution monitoring

• Need
• Fine grain description of operations and message
  parameters.
• Coarse grain description of service and behaviour
  as a whole.

19
Web Services Stack (adapted from IBM)
Ontologies
20
Description Discovery(from IBM)
Wire
Description
Discovery
21
Grid Enabled Optimisation and Design Search
(GEODISE)
Geodise will provide grid-based seamless access
to an intelligent knowledge repository, a
state-of-the-art collection of optimisation and
search tools, industrial strength analysis codes,
and distributed computing data resources
Funded by EPSRC
http//www.geodise.org
22
GEODISE Demo
(1) Security Infrastructure Authentication
Authorisation
(2) Define Geometry to optimise
Nacelle Design
Axisymmetric (2D)
3D
(3) Sample Objective function to build Response
Surface Model
Grid Computing
23
GEODISE Demo
(4) Optimise over Response Surface Model
(5) Grid Database Query and Postprocessing of
Results
Automated Data Archiving
24
Ontologies and Knowledge in GEODISE
Ontology and Knowledge Capture for Design Process
Ontology Driven Service Composition
25
Ontologies and Web Services information and
knowledge integration - Qinetiq

• Combine Web Services and Semantic Web
• Ontologies of services
• Ontologically-informed service parameters
• Build on emergent DAML Services work
• Separate domain-independent components of
  messages from domain-specific parts
• Domain-independent pragmatics
• Domain-specific contents
• Agent Communication Languages
• FIPA ACL Process Ontology

26
Agentifying Web Services
27
Domain Ontology - entities
28
Domain Ontology - events
29
FIPA ACL Ontology - processes
30
Situational Awareness in Humanitarian Relief

• Unseasonal rainfall brings about a rapid rise in
  water levels and widespread flooding in a heavily
  populated river delta.
• Emerging humanitarian crisis requires that
  flood-displaced persons be moved to a place of
  safety and provided with food and shelter.

31
Delta Flood Scenario

• Number of possible data sources
• Media reports
• Meteorological forecasts and reports
• ELINT
• Reports from NGOs
• Other field reports
• Multiple foci of situational awareness
• Refugee concentrations and movements
• Communications and transport infrastructure
• Weather conditions and river levels, current and
  predicted

32
System Architecture
33
Core Knowledge Technologies in this Scenario

• Information Customisation
• Views on data for needs of different users
• Service descriptions used to select sets of
  services to provide data to user
• Information Integration
• Large number of possible data sources
• Integration of data from different services
  presents challenges
• Confirmation of reports
• Certainty
• Provenance
• Timeliness

• Information Extraction
• Newsfeeds from media agencies
• Text-based
• Geospecific references in stories
• Push medium
• Extract salient details
• Time, place, etc
• Confirm against other sources

34
myGrid http//www.mygrid.org.uk
Building personalised extensible environments for
data-intensive in silico experiments in biology
myGrid Middleware
35
myGrid http//www.mygrid.org.uk
Specialises. All concepts are subclassed from
those in the more general ontology.
Upper level ontology
Contributes concepts to form definitions.
Task ontology
Publishing ontology
Informatics ontology
Molecularbiology ontology
Organisationontology
Bioinformatics ontology
Web serviceontology
36
myGrid demo
Portal
Repository Client
Ontology Client
Workflow Client
Meta Data Ontology
Personal Repository
Workflow Repository
Meta Data Service Type Directory
37
myGrid http//www.mygrid.org.uk

• Data intensive Grid for bioinformatics
• Services described using DAMLOIL ontology.
• Services organised, queried and matched using
  subsumption reasoning.
• Descriptions controlled by concept satisfiability
  reasoning.

38
Ontologies Services

• Typing
• Controlling inputs and outputs of services.
• Mapping between WSDL / OGSA XML Schema types to
  (DAMLOIL) concepts

• Classifying
• Indexing services and data
• OGSA factories and service instances.
• Organising services based on reasoning over the
  service descriptions.

39
Capturing Semantics Tuecke

• Service description obviously captures interface
  syntax
• But capturing semantic meaning is critical for
  discovery
• Not only does the service accept an operation
  request with a particular signature
• But it should also respond as expected
• as expected is usually defined offline in
  specifications
• Approach name everything
• Use names as a basis for reasoning about
  semantics
• THIS IS NOT ENOUGH!

40
Semantic Web Services
UDDI
types
XML Schema
businessEntity
messages
businessService
portType
operation
binding Template
binding
service
tModel
WSDL
41
SemanticGrid Services
types
XML Schema
messages
serviceType
Registry
portType
operation
serviceDataDescription
compatibilityAssertion
serviceImplementation
42
SemanticGrid Services
types
XML Schema
messages
serviceType
Registry FindServiceData SubscribeServiceData
portType
operation
serviceDataDescription
serviceType
serviceData
43
Ontology opportunities

• Types and Messages
• ltmessage name findFunctionRequest/gt
• ltpart nameprotein typemyxsdsequencegt/
• /messagegt

44
Ontology opportunities

• portType
• ltportType nameProteinServiceBeangt
• ltoperation name findFunctiongt
• ltinput namefindFunctionRequest
  messagetnsfindFunctionRequest/gt
• ltoutput namefindFunctionResponse
  messagetnsfindFunctionResponse/gt
• lt/operationgt
• lt/portTypegt

45
Service Data Description
Metadata Drawn from an ontology acting as a schema
serviceDataDescription
describes
Data values Drawn from an ontology acting as a
controlled vocabulary for content
serviceData
46
Finding and substituting

• Building registries
• Topology of services used for discovery
• Finding based on property descriptions on Service
  Data
• Matching close enough (c.f. Condor ClassAtt)
• A service that maps an protein sequence to a gene
  product function
• A service that maps an enzyme sequence to a gene
  product function

47
Finding/Matching Grid Services

• FindServiceData
• queryByServiceDataName
• query by properties of service

48
DAML-S http//www.daml.org

• Defines ontologies for the construction of
  service models.

WSDL
UDDI
49
DAML-S Service Profile

• Derive service adverts service requests,
  populating service registries, automated service
  discovery, matching of capability adverts to
  requests

DAML-S
DAMLOIL based types

• Functionality
• High level overview of purpose
• Preconditions requirements
• Inputs Outputs
• Effects of execution
• Levels of quality
• Service provider

Process Model
Inputs/Outputs
Atomic Process
Operation
Message
Binding to SOAP, HTTP etc
WSDL
50
DAML-S Service Models

• Automated service invocation, composition,
  interoperation, coordination, monitoring.
• Capability matching

• More detailed description of operations
• Described in terms of a set of processes
• Preconditions
• Inputs Outputs
• Effects of execution
• Statements about runtime behaviour and
  interaction pattern services through workflow
  constructs sequence, fork

• Service Ontologies
• Ontologies of service types
• Ontologies to augment service requests
• Quality
• Costs
• Security

51
What do ontologies offer?

• Common framework for integration
• OpenMMS, TAMBIS, ONION
• Search support, querying matching
• GO, MGED, UMLS, MeSH
• Intelligent interfaces for queries and data
  capture
• Ingenuity web based products, TAMBIS.
• Control Semantics

52
Integration through shared descriptions using
ontologies

• Conformance to shared controlled vocabularies for
  content.
• Shared semantic model for metadata and data
• Gene Ontology for describing gene function in
  FlyBase, MouseBase, SGD, Swiss-Prot, Interpro
  etc
• UMLS, PharmGKB, MGED

53
From Bertram Ludäscher, SDSC
54
Information integration through mediation

• TAMBIS Transparent Access to Multiple
  Bioinformatics Information Sources.
• Domain ontology global schema
• Sources not visible to users.
• http//img.cs.man.ac.uk/tambis

55
TAMBIS Architecture

• Ontology described using Description Logic.
• Query formulation ontology browsing concept
  construction.
• Reasoning about concept models, answering
  questions like
• What can I say about Proteins?
• What are the parents of concept X?
• Wrapper service Kleisli.

Biological Bioinformatics Ontology
Query Formulation Interface
56
TAMBIS

• Query multiple databases using an ontology as a
  global model

57
Specification of content

• Controlled vocabularies
• Controlled description
• Accurate data collection
• Enhanced and accurate retrieval
• Expectation setting
• Step towards unification and interoperation
• Gene Ontology. MGED, UMLS, Mouse Anatomy etc
• Taxonomy
• Organisational classification
• Index
• Systematic explicitness

58

Knowledge Driven content
Controlled Vocabularies for Describing Alleles
59
Ontology driven forms.

• Clinergy.
• Intelligent interfaces.

60
What do ontologies offer?

• Support for knowledge intensive applications.
• Text extraction.
• Text generation.
• Problem Solving Environments.
• Decision support.
• Control Semantics

PASTA Protein structure extraction from texts to
support the annotation of PDB http//www.dcs.shef.
ac.uk/research/groups/nlp/
61
What do ontologies offer?

• Knowledge repository.
• Knowledge discovery.
• Decision support.
• Hypothesis generation.
• RiboWeb, Ingenuity, PharmGKB, EcoCyc, PlasmoCyc,
  Biopathways, AROM
• Control Semantics Inference

62
Role of Knowledge
Problem solving Environments
Describing, Advertising, Discovering Matching
Services
Intelligent portals
Model and content
Ontologies
Composition Constraining, Description generation
controlled content annotating results
Typing inputs outputs
Typing, controlled content
Metadata
Databases
Workflows
63
http//www.geneontology.org
64
Gene Ontology http//www.geneontology.org

• a dynamic controlled vocabulary that can be
  applied to all eukaryotes
• Built by the community for the community.
• Three organising principles
• Molecular function, Biological process, Cellular
  component
• Isa and Part of taxonomy but not good!
• 10,000 concepts
• Lightweight ontology, Poor semantic rigour.
• Ok when small and used for annotation.
• Obstacle when large, evolving and used for mining.

65
Gene Ontology used for

• Controlled annotation quality consistency
• Mediation navigation through content
• Classification / query / index

InterPro
SWISS-PROT
BLAST
66
Descriptive knowledge

• Integration of structured(?) and semi-structured
  data.

67
Annotation
GO annotations
Gene detail page in MGD for the vitamin D
receptor gene, Vdr
68
Opens browser
69
Search returns children
70
Returns annotated terms
71
(No Transcript)
72
(No Transcript)
73
Gene Ontology as index

• Relies on the Gene Ontologys taxonomy.
• Lack of semantics and rigour begins to be
  problematic.

74
How did GO happen?

• A respected leader with vision and influence who
  got some financial backing.
• A large collection of data resources.
• A motivated community with a problem prepared to
  do the work.
• A culture of curation.
• Early adopters.
• A legacy of controlled vocabularies.
• A desire to do something NOW.

75
Approach Phase 1

• Childhood
• A small team built version 0 practically in their
  spare time.
• Published using primitive tools. No attempt to
  get it right before publication.
• Responsive feedback and curation from the
  community, who feel as they will be listened to
  and it is theirs.
• Used to annotate repositories. One purpose.

76
Approach Phase 2

• Adolescence
• Funded team of curators.
• Larger teams of specialists.
• Regular face-to-face meetings.
• Development and deployment tools and
  environments.
• Open source, simple.
• Methodology for evolution and change
• Depreciation of terms
• Linkage to other Controlled Vocabularies, e.g.
  Swiss-Prot keywords.

77
Approach Phase 3

• Adulthood
• Adoption of better methodologies and tools for
  cleaning up, managing and evolving the
  ontology.
• DAMLOIL
• GONG, GOET projects .
• Different purposes to original intent.

78
Summary

• Building ontologies for the sake of it is
  pointless, but this is sometimes forgotten!
• Use the ontology before it is finished because
  it never will be.
• Lots of work on ontology creation.
• Less work on ontology deployment.
• Most architectures use an ontology server.
• If it is useful it will be used.

79
Further Reading

• ltto be suppliedgt

80
Components in an Ontology Driven Approach
Reasoner
Ontology Language
Lexicon
Ontology Server
Ontologies
Parser