Semantic Web: Customers and Suppliers

1
Semantic WebCustomers and Suppliers

• Rudi Studer
• Institut AIFB, Universität Karlsruhe (TH)
• FZI Forschungszentrum Informatik
• Ontoprise GmbH
• Invited Talk _at_ ISWC2006, Athens, GA, USA
• November 9th, 2006

2
ISWC Looking Back

• ISWC2002 Sardinia, IT
• 95/27 sub/acc, 4 tutorials
• ISWC2003 Sanibel Island, FL, US
• 262/49 sub/acc, 6 workshops, 4 tutorials
• ISWC2004 Hiroshima, JP
• 205/48 sub/acc, 8 workshops, 6 tutorials
• ISWC2005 Galway, IR
• 217/54 sub/acc, 9 workshops, 4 tutorials

3
ISWC Connecting Communities

• 2002 Tutorial on Description Logic (KR)
• 2003 Workshops on Practical and Scalable
  Semantic Systems (DB) and on Human Language
  Technology (NLP) for the Semantic Web and Web
  Services
• 2004 Paper session on Semantic Web Mining (ML)
• 2005 Workshop on Semantic Web Enabled Software
  Engineering (SE)

4
Agenda

• Presence of Semantic Web at Top Events of Other
  Communities
• Customers and Suppliers
• Knowledge Representation (KR)
• Databases (DB)
• Software Engineering (SE)
• Natural Language Processing (NLP)
• Machine Learning (ML)
• Business Aspects
• Trends and Take Home Messages

5
IJCAI, KR

• IJCAI2001 Workshops on Ontology Learning, on
  Ontologies and Information Sharing and on IEEE
  Standard Upper Ontology
• IJCAI2003 Tutorials on Ontologies -
  Representation, Engineering and Applications and
  Ontology-Based Information Integration, Workshop
  on Ontologies and Distributed Systems
• IJCAI2007 Invited talk by Carole Goble on The
  e-Scientist is the Semantic Web's Friend (or a
  Friend Of A Friend), Workshop on Semantic Web for
  Collaborative Knowledge Acquisition
• KR2002 Invited talk by Jim Hendler on The
  Semantic Web KR's Worst Nightmare?, Workshops on
  
• KR2004 Invited talk by Peter Patel-Schneider on
  What Is OWL (and Why Should I Care)?, Workshops
  on
• KR2006 Invited talk by Alon Halevy on
  Dataspaces Co-existence with Heterogeneity

Status Synergetic
6
VLDB, SIGMOD/PODS

• VLDB2001 Invited talk by Pierre-Paul Sondag on
  The Semantic Web Paving the Way to the Knowledge
  Society
• VLDB2003 Tutorial on The Semantic Web Semantics
  for the Data on the Web
• VLDB2004 Invited talk by Alon Halevy on
  Structures, Semantics and Statistics
• Semantic Web and Databases (SWDB) 2003 - 2006 ,
  VLDB workshops in 2005, 2006 on Ontologies-based
  techniques for DataBases and Information Systems
• SIGMOD/PODS2006 Invited talk by Alon Halevy on
  Principles of Dataspace Systems, Invited tutorial
  by Enrico Franconi on The Logic of the Semantic
  Web

Status Knowledgeable
7
ACL, SIGIR, ECML, ICML

• COLING/ACL 2006 Workshop on Ontology Learning
  and Population
• EACL 2006 Tutorial on Ontology Learning from
  Text
• LREC 2006 Invited talk by Enrico Motta on The
  role of language and mining technologies in
  engineering and utilizing the semantic web
• ICML 2005 Tutorial on Machine Learning and the
  Semantic Web
• ECML/PKDD 2004 Workshop on Knowledge Discovery
  and Ontologies
• SIGIR 2003 Workshop on the Semantic Web

Status Informed
8
ICSE, SEKE

• ICSE2004 Tutorial on Software Modeling
  Techniques and the Semantic Web
• Specialized conference SEKE Software Engineering
  and Knowledge Engineering, e.g. sessions on
  Ontologies in 2005, 2006, Workshop on Ontology in
  Action in 2004

Status Aware
9
Customers and Suppliers

• Customers what does Semantic Web research
  deliver to other communities?
• Suppliers what do other communities deliver to
  Semantic Web research?

10
Agenda

• Presence of Semantic Web at Top Events of Other
  Communities
• Customers and Suppliers
• Knowledge Representation (KR)
• Databases (DB)
• Software Engineering (SE)
• Natural Language Processing (NLP)
• Machine Learning (ML)
• Business Aspects
• Trends and Take Home Messages

11
Knowledge Representation (KR)
We establish ontology languages for knowledge
representation on the Web

• We deliver only finest KR formalisms and
  deduction algorithms

12
KR as Customer

• Semantic Web delivers challenges beyond Tweety
• E.g. in application domains such as e-Science

13
KR as Customer

• Initiation of standardization processes
• RDFOWL
• RIF WG

14
KR as Supplier

• Semantic Web requirements on KR
• efficient algorithms
• tractable language fragments
• handling uncertainty and inconsistency
• non-monotonic reasoning
• KR delivers representation formalisms and
  reasoning algorithms
• Description Logic (FaCT, Racer, Pellet, KAON2)
• Logic Programming (Ontobroker)
• Integration of DL/LP (Motik et al., 2004 and
  2006)

15
Integrating DLs and Logic Programming
Acknowledgements Boris Motik, Riccardo Rosati

• DLs are good at
• representing taxonomical knowledge
• representing incomplete information
• unknown individuals and disjunctive knowledge
• But we also want
• to represent arbitrary relationships between
  objects
• represent database-like constraints
• represent exceptions
• Logic programming addresses many of these issues
• Hybrid MKNF knowledge bases
• consist of a DL knowledge base a logic
  program
• are fully compatible with DLs
• are fully compatible with logic programming
• bring together the best of both worlds

16
Databases (DB)
We deliver models for query answering in open,
heterogeneous environments
We deliver efficient management of large data sets
17
Databases vs. Semantic Web

• Databases
• Scalability
• Performance
• Performance
• PerformanceB. Lindsay, IBM Fellow
• Controlled settings
• Closed world assumption

• Semantic Web
• Expressive KR languages
• Description logics
• Uncertainty, heterogeneity and openness of the
  WWW
• Decentralized, ad-hoc settings
• Open world assumption

18
Trends in DatabasesAcknowledgement Alon Halevy

• DB trends
• From DBs to dataspaces
• From integration to co-existence
• Dataspaces Franklin, Halevy, Maier 2005
• pay-as-you-go data management
• Dataspace querying, evolution and reflection
• Need for KR services
• Example Scenarios
• Personal Information Management
• Enterprise Information Integration
• Querying the WWW

19
Example Personal Information Management
Semex, CALO, Haystack
20
DB as Customer Challenge 1 Integration in
Dataspaces

• Formal models for query answering Recent
  developments in the field of knowledge
  representation (and the Semantic Web) offer two
  main benefits as we try to make sense of
  heterogeneous collections of data in a dataspace
  simple but useful formalisms for representing
  ontologies, and the concept of URI (uniform
  resource identifiers) as a mechanism for
  referring to global constants on which there
  exists some agreement among multiple data
  providers."
• Abiteboul et al., The Lowell Database Research
  Self-Assessment, CACM May 2005/Vol. 48, No. 5
• Semantic integration of data sources
• Integration of structured and unstructured data
• Ranking of answers

21
DB as Customer Challenge 2 Semantic Mappings

• Methods for answering queries from multiple
  sources without set of pre-defined correct
  semantic mappings
• A semantic heterogeneity solution capable of
  deployment at Web scale remains elusive. The
  same problem is being investigated in the context
  of the Semantic Web. Collaboration between groups
  working on these and other related problems, both
  inside and outside the database community, is
  important.
• Abiteboul et al., The Lowell Database Research
  Self-Assessment, CACM May 2005/Vol. 48, No. 5
• Approximate mappings
• Measuring the accuracy of mappings
• Emergent Semantics Gossip based algorithms
• Infer mappings, reasoning about mappings

22
DB as Customer Challenge 3Uncertainty and
Inconsistency

• Life is imperfect with dataspaces
• Semantic relationships are uncertain
• Data sources may be imprecise
• Data will often be inconsistent
• Reasoning with inconsistent knowledge
• Diagnosis and repair, belief revision
• Languages for modeling fuzzy and uncertain
  knowledgeWorkshop on Uncertainty Reasoning for
  the Semantic Web (URSW _at_ ISWC2005,2006)

23
DB as Supplier Deductive Database Techniques in
KAON2

• Efficient reasoning with large datasets (ABox) is
  hard with standard methods for OWL reasoners
  (tableaux algorithms)
• Deductive databases can efficiently handle large
  data quantities
• Idea apply techniques from the field of
  (disjunctive) deductive databases
• join-order optimization
• magic sets optimization

KB ? if and only if DD(KB) ? for a
ground fact ?
DL knowledge base KB
Query
Disjunctive datalog program DD(KB)
24
Software Engineering
We deliver formal models which enable e.g.
reasoning about resources
We deliver architectures, tool support, visual
modelling techniques
25
Ontologies vs. ModelsAcknowledgements Colin
Atkinson

• Ontologies
• originated from the artificial intelligence world
  for the purpose of precisely structuring
  knowledge
• new knowledge derived by automated reasoning
• characterized by OWL as the flagship language
• formal semantics (description logic)

• Models (à la MDA)
• originated from the software engineering world
  for structuring the specification of software,
  abstracting from platform specific aspects
• information defined prescriptively for
  construction
• characterized by UML as the flagship language
• semi-formal semantics (metamodels)

?
26
Ontology Definition Metamodel
Metaobject Facility (MOF)
Metaobject Facility (MOF)
Meta- Meta- Model
UML Profile (Visual Syntax)
Ontology Metamodel
Meta- Model
Mappings
UML 2.0 Model
Ontology
Model
27
SE as SupplierMOF-based Ontology Development

• MDA enables interoperability
• MDA-based tool support (modeling tools, model
  management)
• Independence of specific formalisms
• Definition of the ontology model in an abstract
  form, independent of the particularities of
  specific logical formalisms
• Language mappings (groundings) define the
  transformation to particular formalisms
• Reuse of UML for visual modeling
• see NeOn approach for networked ontology model

28
SE as Customer MOF and Semantic Web
Acknowledgements Elisa F. Kendall

• MOF technology streamlines the mechanics of
  managing models and model transformation
• Semantic Web technologies provide reasoning about
  resources
• Semantic alignment among differing vocabularies
  and nomenclatures
• Consistency checking and model validation, e.g.
  business rule analysis
• Ask questions over multiple resources that one
  could not answer previously
• Policy-driven applications to leverage existing
  knowledge and policies
• Example A Formal Framework for Reasoning on UML
  Class Diagrams Lenzerini et al. 2002

29
Natural Language Processing (NLP)
We deliver all kinds of ontologies and reasoning
support, e.g. to improve disambiguation
We deliver solid information extraction methods
and tools
30
NLP as Customer

• Domain ontologies for disambiguation, e.g.
• Compound interpretation (see OntoQuery project)
• Lexical Ambiguities
• corner has 11(!) meanings (synsets) in Wordnet
• but in specific domains much less meanings are
  typically relevant, e.g. in the soccer domain
  (SmartWeb)
• corner as location on the playing ground
• corner as a player action
• Syntactic ambiguities (PP-attachment, )

31
NLP as Customer

• Foundational ontologies for capturing
  domain-independent aspects of meaning
• see Cimiano and Reyle 2006
• Spatial and temporal ontologies to support NL
  interpretation by reasoning

32
NLP as Supplier

• Many methods for information extraction (IE) from
  text have been developed in the past
• see Message Understanding Conferences (MUC)
• Use the Web as a corpus of evidence
• A-Box (PANKOW Cimiano et al. 2004, KnowItAll
  Etzioni et al. 2004)
• T-Box (synonym discovery Turney 2001)
• Automating ontology evaluation
• e.g. w.r.t OntoClean (see AEON Völker et al.
  2005)

33
NLP for Ontology Evaluation
Ahh and how do I evaluate the ontology?

• Understanding OntoClean requires (at least )
  philosophical, modelling and particular domain
  knowledge
• Even for experts applying OntoClean is tedious
  and time-consuming
• Automatic Evaluation of ONtologies (AEON)
  facilitates tagging wrt OntoClean meta-properties
• Nature of concepts reflected by human language
  and what is said about instances of these
  concepts
• He is no longer a student. (student not rigid)
• Connecting more than two computers requires a
  hub. (computer is countable thus carries
  identity)
• Pattern-based approach
• Detect positive and negative evidence for
  meta-properties
• Use WWW as corpus Overcome data-sparseness

34
AEON Architecture
AEON
R -I ..
Output Tagged Ontology
Input Ontology
35
Machine Learning (ML)
We improve bag-of-words models with semantics

• We deliver learned
• A-Boxes and T-Boxes

36
ML as Customer

• Inclusion of semantics in bag-of-word models
• Text clustering and classification (Bloehdorn et
  al. 2005)
• Information Retrieval (Gonzalo et al. 1999)
• Semantics in image recognition
• Fuse information from
• different classifiers
• (see ACEMEDIA Project)

37
ML as Supplier

• Use of machine learning methods for A-Box and
  T-Box learning
• Inductive Logic Programming (ILP) for induction
  of concept definitions, e.g. for restructuring
  concept hierarchies (Esposito et al. 2004)
• Discover new associations between concepts (e.g.
  via association rules) (Maedche Staab 2000)
• Learning Taxonomies by
• unsupervised clustering techniques,
• e.g. OntoGen (Grobelnik et al., 2006)

38
Disclaimer

• Other important areas which I could not mention
  here include Agents, Blogs, Grids, Peer-to-Peer
  Systems, Social Networks, Web Services,

39
Agenda

• Presence of Semantic Web at Top Events of Other
  Communities
• Customers and Suppliers
• Knowledge Representation (KR)
• Databases (DB)
• Software Engineering (SE)
• Natural Language Processing (NLP)
• Machine Learning (ML)
• Business Aspects
• Trends and Take Home Messages

40
Business Aspects of Semantic Technologies
41
The Semantic Technology Market Offers High Growth
Potential

• Application areas drive the market for semantic
  technologies
• Enterprise Information Integration (EII)
• Enterprise Content Management (ECM)
• Enterprise Resource Planning (ERP)
• Product Lifecycle Management (PLM)

NeOn Reference Architecture
42
Market Estimation
Semantic Access Integration market Semantic Access Integration market Semantic Access Integration market

  In mio. 2006 2007 2008 2009 2010   CAGR

ECM Worldwide 2.754 3.277 3.900 4.641 5.523 19,00
Semantic ECM () 5 7 10 15 20
  Semantic ECM 138 229 390 696 1105    
EII Worldwide 4.580 5.985 7.116 8.461 10.060 18,90
Semant. Info. Integration () 5 10 15 20 25
  Semant. Info. Integration 229 599 1067 1692 2515    
ERP Worldwide 17.470 18.309 19.187 20.108 21.074 4,80
Semantic ERP () 0 0,5 2 10 20
  Semantic ERP 0 92 384 2011 4215    
PLM Worldwide 6.600 7.920 9.108 10.474 12.045 15,00
Semantic PLM () 5 8 8 8 8
  Semantic PLM 330 634 729 838 964    
Semantic Access Integration 697 1.553 2.570 5.237 8.798    
43
Information Integrator
Views
Business Ontology
Declarative Mappings
Ontologies of Sources
Automated Mapping
ltarticlegt ltarticleidgta-5634lt/articleidgt
ltcategorygtprinterlt/categorygt
ltnamegthp81lt/namegt ltprice
currencyUSDgt500lt/pricegt
ltproducergthplt/producergt ltresolutiongt1960
dpilt/resolutiongt lttypegtlaserlt/typegt . lt/ar
ticlegt
Heterogeneous Sources
44
A Look at REAL CustomersAcknowledgement Richard
Benjamins, iSOCO

• Ontologies are the key differentiating feature of
  Semantic Web technologies
• Semantic integration of heterogeneous sources
• Automatic processing of unstructured information
• One of the current main obstacles for Semantic
  Web technologies is the need for Ontologies
• They are hard to construct and maintain
• May involve many stakeholders
• Their costs are difficult to estimate and control
• Before Semantic Web technology goes to mainstream
  market, potential customers (businesses and
  governments) need to perceive that ontologies are
  
• Doable, controllable and manageable, affordable
• An asset for creating/maintaining competitive
  advantage
• An asset that can be sold as high-value content
• Understanding and controlling cost factors of
  ontology engineering is critical (see OntoCom
  Paslaru et al., 2006)

45
Agenda

• Presence of Semantic Web at Top Events of Other
  Communities
• Customers and Suppliers
• Knowledge Representation (KR)
• Databases (DB)
• Software Engineering (SE)
• Natural Language Processing (NLP)
• Machine Learning (ML)
• Business Aspects
• Trends and Take Home Messages

46
Take Home Messages

• KR Synergetic collaboration
• DB Similar strategic challenges and goals
• NLPML Huge potential, not yet exploited
• SE Potential recognized, but still in early
  stage
• Business Aspects Existing and growing market for
  Corporate Semantic Web applications

47
Trends
48
Trends

• Web Science (Berners-Lee et al.)
• studies the scientific, technical and social
  challenges underlying the growth of the Web
• Semantic Web as important building block
• Convergence Web 2.0 and Semantic Web
• Web 2.0 Collaborative development of content,
  community effects ? The Social Web
• Semantic Web structuring principles,
  well-defined and reusable meaning for metadata,
  mash ups on the fly

49
Semantic MediaWiki

• Enable wiki authors to structure information
• RDF export of this structure
• Knowledge reusable inside the wiki

• Typed links
• ISWC2006 is in Athens, GA
• ISWC2006 is in locationAthens, GA
• Typed Attributes
• ISWC2006 starts November 7
• ? startsNovember 7, 2006
• Query for all US conferences in autumn 2006

starts
50
Semantic MediaWiki

• Collaborative management of semantically enriched
  content, tailored towards usability and
  simplicity
• Edit Annotate annotation easy as wiki-editing
  unconstrained, collaborative,
  version-controlled
• Search Explore semantic search, novel
  browsing, and easier maintenance as instant
  rewards
• Share Reuse content exported as browsable OWL
  DL/RDF, reusing existing vocabularies and
  ontologies
• Thousands of real users in many languages
• Semantics to the people!

http//ontoworld.org/wiki/Semantic_MediaWiki
51

• Thank You!
• Rudi Studer
• Institut AIFB, Universität Karlsruhe (TH)
• http//www.aifb.uni-karlsruhe.de/
• with contributions from Philipp Cimiano, Peter
  Haase, Pascal Hitzler, Markus Krötzsch,
  Hans-Peter Schnurr, York Sure, Denny Vrandecic
  Semantic Karlsruhe