Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science...

95
Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov , Dmitriy Zheleznyakov KRDB research centre, Free University of Bozen-Bolzano, Italy

Transcript of Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science...

Page 1: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

Evolution of OWL 2 QL and EL Ontologies

Bernardo Cuenca Grau, Ernesto Jiménez-RuizComputer Science Department, University of Oxford, UK

Evgeny Kharlamov, Dmitriy ZheleznyakovKRDB research centre, Free University of Bozen-Bolzano, Italy

Page 2: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

2

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 3: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

3

Ontologies: schema + data

o Schema provideo standard vocabularies for datao a way to structure datao means for machines

to be able to understand data

o Schemas are in terms ofo classes:

Person, Country, ...o (binary) properties:

State-of-Origin, Subclass-of, ...

o Data is a collections of factso Instantiations of classes o Instantiations of properties

Page 4: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

4

Domain ontologies

o Goal: to provide standard vocabularies to communities

o Clinical sciences ontologies:o SNOMED CT: Systematized Nomenclature of Medicine - Clinical

Terms o > 311k concepts

o NCIt: National Cancer Institute thesauruso ~ 89k concepts, 200m cross links between them [NCI]

o FMA: Foundational Model of Anatomyo 75k classes, 168 relations, 120k terms, 3.1m relat. inst.

Page 5: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

5

Languages for domain ontologies

o Domain ontologies areo complex and largeo manually createdo should be error free

o Languages that are natural for domain ontologieso flexible to capture complex interactiono logic-based (e.g., based on Description Logics)

o Ontology Web Language: OWL 2o OWL DLo OWL 2 QLo OWL 2 ELo e.g. SMOMED

forall x: instance-of (x, Common cold) exists y: instance-of (y, Virus) and causative-agent (y, x)

Page 6: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

6

Evolution in SNOMED

o Development teamso 1 main team ando 4 geographically distributed teams o each team makes modifications

o Every 2 weeks the main teamo integrates changes, resolve

conflicts

o From 2002 to 08 SNOMED went from 278k to 311k concepts [SM-1]

o Example of modifications:o In Jan. 2006 a number of

concepts from the “Clinical finding” hierarchy were moved to the “Event hierarchy” [SM-2]

Page 7: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

7

Evolution in NCI and FMA

o Developers of NCI do over 900 monthly changes [HKR’08]

o 20 full time editors for NCIo they work o independently o on a separate copy of the ontology

o There is one curator for NCIo every 2 weeks curator o reviews changes using a workflow management toolo approves the changes

o they merge results once a montho there is one curator who curates once a month

o FMA “is an evolving computer-based knowledge source ...” [FMA]

Page 8: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

8

Evolution of domain ontologies

o Evolution of domain ontologies is common

o Ontologies are changed by o insertion of axioms o deletion of axioms

o Evolution affects botho schema level o data level

Evolution of domain ontologies should be error free

Page 9: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

9

Design errors: incoherency

o incoherency is a schema level design error: o incoherent concept = empty concepts o can be caused by disjointness and cardinality restrictions

o incoherent role = empty roleo can be caused by disjointness and cardinality restrictions

EquivalentClasses( :Nothing ObjectIntersectionOf(  :Airplane :Boat))

SubClassOf( :Amphibian :Airplane)SubClassOf( :Amphibian :Boat )

Page 10: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

10

Design errors: inconsistency

o Inconsistency is an error that involves both o data level and o schema level

o Inconsistency:o disjoint concepts are Instantiatedo functionality is violatedo number restrictions are not respected

EquivalentClasses( :Nothing ObjectIntersectionOf(  :Airplane :Boat ))

ClassAssertion(:Airplane :BerievA-40 ) ClassAssertion(:Boat  :BerievA-40 )

Page 11: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

11

Insertions bring errors

o Insertions introduce errors which should be repaired

o Incoherency

o Inconsistency

o Challenge: how to repair the ontology after “bad” insertions?

EquivalentClasses( :Nothing ObjectIntersectionOf(  :Airplane :Boat ))

ClassAssertion(:Airplane :BerievA-40 )

EquivalentClasses( :Nothing ObjectIntersectionOf(  :Airplane :Boat ))

SubClassOf( :Amphibian :Airplane)

SubClassOf( :Amphibian :Boat )

ClassAssertion(:Boat  :BerievA-40 )

Page 12: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

12

Deletions bring headache

o Deletions do not introduce (design) errorso no inconsistencyo no incoherency

o Contraction can provoke o restoring of implicit

datao deletion of implicitly

related data

SubClassOf( :Airplane :Transport )

ClassAssertion( :Transport :BerievA-40 )

SubClassOf( :Airplane :Transport ) ClassAssertion( :Airplane  :BerievA-40)

Page 13: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

13

Deletions bring headache

o Deletions do not introduce (design) errorso no inconsistencyo no incoherency

o Contraction can provoke o restoring of implicit

datao deletion of implicitly

related data

SubClassOf( :Airplane :Transport )

SubClassOf( :Airplane :Transport )

ClassAssertion( :Airplane  :BerievA-40 ) ClassAssertion( :Transport :BerievA-40 )

o Challenge: how to respect implicit relations while deleting knowledge?

Page 14: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

14

SPARQL 1.1 Update

o Proposed by HP and based on SPARUL extension of SPARQL for o addingo deletingo updating

RDF triples

o Deletion without deletion effecto only explicit occurrences

of triples are deletedo there is no validation

whether the tupleis still there implicitly

SubClassOf( :Airplane :Transport )

ClassAssertion( :Airplane  :BerievA-40 )

SubClassOf( :Airplane :Transport )

ClassAssertion( :Airplane  :BerievA-40 ) ClassAssertion( :Transport :BerievA-40 )

Page 15: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

15

Syntactic approaches to evolution

o In the ontology:o “Children are baklava fans”o “Children are not cats”

o To delete: “Children are baklava fans”

o To this end it is enough to delete

[HS’05] [JRCGHB’11]

[KPSCG’06]and

o In the resulted ontology:o “Children are not baklava fans”o “Children are not cats” is lost

OK

Not desirable

Page 16: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

16

Semantic approaches to evolution

o How to restore knowledge which o was semantically deleted and o is desirable

o One has to find semantic difference between o the original and o the obtained ontology

o There is a number of approaches and tools to find semantic differenceo Collaborative Protegeo DOGMA-MESSo Content CVS approacho ....

[FDCM’08] [MDM’06]

[JRCGHB’11]

Page 17: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

17

Limitations of current sem. approaches

o Quite application and language oriented

o Heuristic based

o What is missing: the big pictureo a general understanding of evolution of logic based ontologieso proper theory that explains relationships among o different types of ontology modificationso different ontology languages o feasibility and complexity of evolution computation

o There are several attempts to understand logic based evolution

o We are working on that too

2nd part of this tutorial is about current achievements in this direction!

Page 18: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

18

Summary on domain ontologies

o Domain ontologies are o largeo logic based

o Changes in domain ontologieso are frequento are about insertion and deletions

o Insertions easily introduce errors o incoherencyo inconsistency

o Deletions o do not introduce (logical) errors o not trivial: implicit knowledge relationships should be traced

Page 19: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

19

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 20: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

20

Web knowledge bases (ontologies)

o Goal: gathering general purpose knowledge from the Web

o DBpedia:o structural counterpart of Wikipediao 320 classes, 1.650 different properties, 19m facts

o Yago: o combines Wikipedia and WordNet, GeoNames, o 10m entities, 120m facts about them

o (Open)Cyc: o started in 1984, formalizing knowledge manually o logic based KB with reasoingo 47.000 concepts, 306.000 facts

o These ontologies are not statico they constantly change, since Wiki does soo Yago crawls Wikipedia every couple of weeks

...

Page 21: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

21

Languages for Web KBs

o Web KBso have rather simple and small schemaso should be error freeo errors are rare

o Languages that are natural for domain ontologieso able to describe basic thingso SubClassOf, Domain, Range, etc.

o These languages are:o Resource Description Framework with Schema: RDF and RDFSo a bit of OWL 2: owl:equivalentClasso Some rule languages: OWL 2 RL

o Evolution is performed ad hoco Each KB has its approach

Page 22: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

22

Evolution in DBpedia

o DBpediao 18 functional propertieso new information is obtained from Wikipediao new data can violate functional properties

o Inconsistency is possible

FunctionalObjectProperty( :netIncome)FunctionalObjectProperty( :co2Emission)FunctionalObjectProperty( :height) ...

Page 23: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

23

Evolution in Yago

o Yago is a clean (inconsistency fee) ontologyo 95% of accuracy - manually validated on 6k factso New knowledge should not cause contradictions

Page 24: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

24

Yago consistency check [Yago-1]

o Yago has rules to check consistencyo check uniqueness of entities and functional argumentso domains and rages of relationso type checking

Rock Singer

type

1935born

Singer

subclassOf

subclassOf

Physics born

Guitarist

Guitar

Page 25: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

25

Summary on Web KBs

o Web KBs aim at consistency

o Schemas of Web KBs are rather simple and smallo it is hard to make errors

o Evolution is performed ad hoc

Page 26: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

26

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 27: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

27

Ontologies for semantic markup

o Goal: o to nest semantics within existing content on web pageso to help search engines, crawlers and browsers find the right data

Person:• name• photo• URL• ...

text

embedding semantic annotations

Page 28: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

28

Standards for semantic markup

o Microformats, since 2003 o Small set of fixed formats. E.g.:o hcard : people, companies, organizations, and places o XFN : relationships between peopleo hCalendar : calendaring and events

o RDFa: Resource Description Framework – in – attributeso since in 2004, W3C recommendationo serialization format for embedding RDF data into HTML pages o can be used together with any vocabulary, e.g. FOAF

o Microdatao alternative techniques for embedding strucuted datao proposed in 2009, comes with HTML 5

Page 29: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

29

Is semantic markup popular? [CB’12]

o Yahoo Crawl of 2011o 12 billion pages were crawledo 431 million of then contain RDFa

in 2011 - 3.5% of the HTML pages had structured (meta) data

Vocabulary Number of web sites

Dublin Core 344.545

Open Graph Protocol 177.761

Creative Commons 37.890

Google’s Rich Snippets Vocab. 6.083

Friend-of-a-Friend 2.545

... ...

Page 30: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

30

Big step in promoting ontologies

o Schema.org initiative: o started on June 2011 o initiated by Bing, Google, Yahoo!, Yandexo they propose:

to mark up / annotate websites with metadatao they support: Microdata

Page 31: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

31

Schema.org ontologies

o Metadata by Schema.org:o Persono Organizationo Evento Placeo Producto ...

o 200+ types

Page 32: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

32

Where can you see Scmeha.org impact?

Page 33: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

33

Semantic markup today

o Common Crawl foundationo goal: building and maintaining an open crawl of the Webo current data is about 5 billion web pages

o WebDataCommons.org projecto goal: extracting Microformats, Microdata,

RDFa from Common Crawl corpuso Feb 2012: o processed 1.4 billion HTML pages of CC corpus o 20.9 Terabyte of compressed datao this is a big fraction of the Web

Page 34: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

34

Structured Web data is fast growing

o 1.4 billion HTML pages processes

o 188 millions of them contain structural datain Microformat, Microdata, RDFa [CB’12]

o This data is 3.2 billions RDF triples

13% of the HTML pages contain structured (meta) data

from 2011 to 2012 the fraction of structured data went from3.5% to 13%

Page 35: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

35

Evolution at schema level: Schema.org

o It is a very simple and coherent schema

o Coherency o basic Schema.org vocabulary can be mapped to RDFSo RDFS schemas are always coherent so does Schema.org

o What is used from RDFS: [SO-2]

o subclasso domain, range restriction of propertieso literal, o ...

o Schema can be extendedo mechanism: specialization o of classes, properties, enums

o Person/Engineer [SO-3]

PloiceStation A police station. Subclass of: CivicStructure Subclass of: EmergencyService

creator The creator/author of this Creative Work Domain: CreativeWork Domain: UserComments Range: Person Range: Organization

Page 36: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

36

Evolution at data level: Schema.org

o It is RDFS embeddable no inconsistency is possible

o Schema.org convention: on range restriction

[SO-1]

o each property may have 1 or more types at its rangeo the value(s) of the property should be instances

of at least one of these types

o Thus, they accept that data can be inconsistent

Page 37: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

37

Evolution at data level: Schema.org

o Is data inconsistency important?

o Data gathered by crawling the Web is inconsistent by natureo data consistency is not important o data consistency is unrealistic

o Data maintained locally can be consistent o consistency of data can be important

In the spirit of "some data is better than none",

we will accept this [inconsistent] markup and do the best we can.

[SO-1]

Page 38: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

38

Summary on semantic markup

o Semantic mark up schemas areo smallo very simple

o In many cases logical errors with semantic markup are simply impossible

o Consistency and coherency is in general not important

Page 39: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

39

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 40: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

40

Summary: ontologies and evolution

o Three major groups of ontologieso unification of terminology by specific communities o domain ontologies

o storing general purpose web content in o Web knowledge bases

o enriching Web content with information understandable by agents, e.g. crawlers – 13% of Web data is enriched!o ontologies for semantic markup

o In all these cases ontologies are dynamic o insertions and o deletions

happen at the level of o schema ando data

Page 41: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

41

Summary: attitude to evolution

ontologies for semantic markup

domain ontologies

Web knowledge bases

o schema is simple (RDFS): errors are (almost) impossible

o data may disrespect the schema o “some data is better than none”o “do the best we can”

o schema is more involved but still no incoherency(RDFS + some OWL e.g., functionality)

o data may be inconsistento conflicts can be detected by simple reasoning o many problems are solved by type checking

o schema is complex (OWL 2) – incoherencyo data can easily be inconsistento coherency + consistency are vitalo logical reasoning can guarantee it

don’tcare

logic based

Page 42: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

42

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 43: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

43

Logic-Based Evolution

o The main principle of logic-based evolution isthe principal of minimal changeo Ontologies should change as little as possible

o There are two main classes of logic-based approaches:o Model-based approach (MBA)

o Formula-based approach (FBA)

o There are two main types of evolution:o Update (or revision), when new information is addedo Contraction (or erasure), when some old information is retracted

o We illustrate o update with MBA o contraction with FBA

[KM’91][EG’92]

[LLMW’06][QD’09][WWT’10]

[Wins’90]

[CKNZ’10]

[KZ’11]

[Wins’90]

[DGLPR’09]

Page 44: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

44

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 45: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

45

MBA: Evolution Process

Ontology Models

ModelTransforme

r

Evolvedmodels

Evolvedontology

Newdata

Page 46: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

46

MBA: Ontology to Models

Priest

Bob

Wife Husband

Mary John

Model 1:

Model 2:

Priest

Bob

Adam

Wife Husband

Mary Peter

Page 47: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

47

MBA: Evolution Process

Ontology Models

ModelTransforme

r

Evolvedmodels

Evolvedontology

Newdata

Page 48: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

48

MBA: Data Evolution

Priest

Bob

Wife Husband

Mary John

Model 1:

Model 2:

Priest

Bob

Adam

Wife Husband

Mary Peter

…Dalal’s operator

Satoh’s operator

Winslett’s operator

Page 49: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

49

Winslett’s operator

Wife Husband

Mary John

Priest

Bob

Model 1:

Priest

Bob

Adam

Model 2.1:

Wife Husband

Mary John

Priest

Bob

Adam

Wife Husband

Mary Peter

Anna John

Model 2.2:

Wife Husband

Mary Peter

✘ Model 2:

MBA: Data Evolution

Page 50: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

50

MBA: Evolution Process

Ontology Models

ModelTransforme

r

Evolvedmodels

Evolvedontology

Newdata

Page 51: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

51

Wife Husband

Mary John

Priest

Bob

Model 1:

Priest

Bob

Adam

Model 2.1:

Wife Husband

Mary John

Priest

Bob

Adam

Wife Husband

Mary Peter

Anna John

Model 2.2:

MBA: Models to Ontology

Page 52: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

52

Models

ModelTransforme

r

Newdata

Winslett’s operator

Evolvedmodels

MBA: Issues

Ontology

Evolvedontology

Infinite number?

Infinite number

?

Can be computationally

hard

Can be inexpressible

Page 53: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

53

MBA: Example of Issues

Mary John

Models:

Priest

Bob1.

Wife Husband

Mary Peter

Priest

Bob

John

2.

Wife Husband

Mary Bob

Priest

John3.

Wife Husband

John

Page 54: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

54

MBA: Example of Issues

o The case analysis of the situation with Mary gives that: o Bob is not a priest iff Bob is Mary’s husbando Mary can have as many husbands as she wants,

but if Bob is her husband, she cannot be married to anyone elseo …

o These properties cannot be expressedo in OWL 2 QL noro in OWL 2 EL

[KZ’11]

Page 55: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

55

MBA: Issues with Data Evolution [KZ’11]

o Both OWL 2 QL and OWL 2 EL are not closed under data evolution

o The source of the problem: DisjointClasses

o With some restriction on DisjointClasses:o can be computed in polynomial time

o Without restrictions on DisjointClasses:o can be computed only in OWL DL in exponential timeo can be approximated in polynomial time

(with significant loss of information)

o Conclusion: MBS are not very suitable for ontology evolution

Page 56: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

56

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 57: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

57

FBA: Evolution Process

Ontology

Filter

Evolvedontology

Newdata

Closure

Evolvedclosure

Page 58: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

58

FBA: Ontology to Closure

Page 59: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

59

FBA: Ontology to Closure

Page 60: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

60

FBA: Ontology to Closure

Page 61: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

61

FBA: Ontology to Closure

Page 62: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

62

FBA: Ontology to Closure

Page 63: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

63

FBA: Evolution Process

Ontology

Filter

Evolvedontology

Newdata

Closure

Evolvedclosure

Page 64: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

64

FBA: Schema Evolution

…WIDTIO semantics

Cross-product semantics

Bold semantics

Page 65: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

65

FBA: Schema Evolution

Bold semantics

✘ or ✔?✔ ✘ or ✔?✔ ✘ or ✔?✔

✘ or ✔?✘

?

Page 66: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

66

Bold semantics

✔ ✔

FBA: Schema Evolution

?

Page 67: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

67

FBA: Evolution Process

Ontology

Filter

Evolvedontology

Newdata

Closure

Evolvedclosure

Page 68: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

68

FBA: Closure to Ontology

✔ ✔

✔ ✔

Page 69: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

69

FBA: Evolution Process

Ontology

Filter

Evolvedontology

Newdata

Closure

Evolvedclosure

Bold semantics

Infinite number?

Infinite number

?

Can be computationally

hard

Can be inexpressible

Page 70: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

70

FBA: Example of Issues

✘ or ✔? ✘ or ✔?✔?

Page 71: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

71

FBA: Example of Issues

…✔

✔ ✘?

Page 72: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

72

FBA: Example of Issues

Ontology

Filter

Evolvedontology

Newdata

Closure

Evolvedclosure

Page 73: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

73

FBA: Example of Issues

Page 74: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

74

FBA: Example of Issues

o We cannot keep that gourmets are French,while we have to keep too much useless information:o Gourmets who like bikes are Frencho Gourmets who like those who like bikes are Frencho …

o This cannot be expressed in OWL 2 EL [CJKZ’12]

Page 75: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

75

FBA: Issues of Evolution

o OWL 2 EL is not closed under schema nor data evolutiono The closure is infinite Cannot be captured in general

o OWL 2 QL is closed under both schema and data evolutiono The closure is always infinite The evolved ontology always existso Can be computed in polynomial time

[CJKZ’12]

[CKNZ’10]

Page 76: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

76

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 77: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

77

Syntactic-Deductive Approach

o There are two extreme caseso FBS — it preserves too much informationo In the example, all that information about who likes bikes

o SA — it preserves not enough information.o When it deletes something, does not restore meaningful

entailmentso Deletion without deletion (SPARQL 1.1)

o Possible solution: o to be between these extreme cases,

to preserve only some part of closure.o How big that part is — depends on an application

o This approach is called Syntactic-Deductive (SDA)

[CJKZ’12]

Page 78: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

78

Syntactic-Deductive Approach

SA

FBS

Page 79: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

79

SDA: Example of Issues

…✔

Restriction:n ≤ 2

?✔ ✘

Page 80: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

80

SDS: Example

Ontology

Filter

Evolvedontology

Newdata

Closure

Evolvedclosure

Page 81: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

81

SDS: Example

Page 82: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

82

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 83: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

83

Experiments: Setting [CJKZ’12]

o A contraction SDA algorithm was implemented

o Experimentso Ontology — a fragment of SNOMED with 6802 classeso Reasoner — HermiT [MSH2009]o Facility for computing justification — OWL API [Kal2007]o Closure — only “active” classes, i.e., mentioned in the ontologylogy

Mentioned

Not mentioned

Page 84: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

84

Experiments: Setting [CJKZ’12]

o A contraction SDA algorithm was implemented

o Experimentso Ontology — a fragment of SNOMED with 6802 classeso Reasoner — HermiT [MSH2009]o Facility for computing justification — OWL API [Kal2007]o Closure — only “active” classes, i.e., mentioned in the ontology

Not mentioned

Page 85: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

85

Experiments: Results [CJKZ’12]

# of contracted

axioms

# of recovered axioms

(max/avg/min)

Time (s)(avg)

# of tests

1 52/5/0 135 52

2 96/24/0 217 51

3 195/70/0 176 51

4 257/138/26 169 39

5 281/162/75 165 42

o 95% of recovered axioms are not redundanto Hence, SA leads to significant loss of information

o Average time 2-4 minuteso Time does not depend on the amount of recovered axioms

Page 86: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

86

Experiments: Summary

o SDA contraction is feasibleo Running time is practicalo Approach is scalable

o The most of recovered axioms are logically not redundant

o Recovered axioms seem practically relevant

o We are working on further implementations

Page 87: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

87

Outline

1. Ontologies and evolutiono Domain ontologieso Web knowledge baseso Semantic markup

2. Logic-based approacheso Model-Based approacheso Formula-Based approacheso Syntactic-deductive approacho Experiments

3. Conclusion and directions

Page 88: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

88

Conclusion

ontologies for semantic markup

domain ontologies

Web knowledge bases

don’tcare

logic based

simpleschema

complexschema

o There are three main classes of ontologies

o Ontologies are naturally dynamic o Understanding how to do evolution is important

o Keeping ontologies error free is important for some applicationso The more schema involved the more consistency is important

Page 89: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

89

Conclusion

ontologies for semantic markup

domain ontologies

Web knowledge bases

don’tcare

logic based

simpleschema

complexschema

o Logical based approaches help to prevent errors in ontologieso Model-based approaches o misbehave badly: inexpressibility is built-in

o Formula-based approacheso better, but have issues:

ignorant to original structure, can be impracticalo Syntactic-Deductive Approaches: very promising

Page 90: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

90

Directions

o Proper ontology update language o SPARQL 1.1 Update does not do the right job for many applications

o For domain ontologieso Can one find “the” logic approach?o SDA is one approach – are there better ones?

o Good ontology CVS system

o For web knowledge baseso Can one do more than just ignoring conflicting data?o Maybe some probabilistic techniques can be useful? o How can reliability of knowledge sources help in better “ignoring”?

o For semantic markupso Consistency is impossible o consistent query answering over inconsistent knowledge

Page 91: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

91

References

o [HKR’08] Hartung, M.; Kirsten, T.; and Rahm, E. 2008. Analyzing the evolution of life science ontologies and mappings. In Proc. of DILS, 11–27.

o [SM] Spackman K. SNOMED RT and SNOMEDCT. Promise of an international clinical terminology. MD Comput. 2000 Nov;17(6):29.

o [SM-1] http://www.ihtsdo.org/snomed-ct/snomed-ct0/adoption-of-snomed-ct/

o [SM-2] http://www.ihtsdo.org/fileadmin/user_upload/doc/download/doc_UserGuide_Current-en-US_INT_20120131.pdf

o [FMA] http://sig.biostr.washington.edu/projects/fm/AboutFM.html

o [CB’12] Christian Bizer: Topology of the Web of Data. Joined keynote talk at the 2nd Workshop on Linked Web Data Management (LWDM2012) and the 3rd Workshop on Business Intelligence and the Web (BEWEB2012), Berlin, Germany, March, 2012.

o [SO-1] http://schema.org/docs/datamodel.html

Page 92: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

92

References

o [SO-2] http://schema.org/docs/schema_org_rdfa.html

o [SO-2] http://www.schema.org/docs/extension.html

o [SO-3] http://schema.rdfs.org/mappings.html

o [SO-4] https://raw.github.com/dcmi/schema.org/master/mappings_schema.org.xml

o [NCI] https://wiki.nci.nih.gov/display/EVS/NCI+Thesaurus+versus+NCI+Metathesaurus

o [Yago-1] Hady W. Lauw, Ralf Schenkel, Fabian M. Suchanek, Martin Theobald, Gerhard Weikum, "Harvesting Knowledge from Web Data and Text” Tutorial at the 19th International Conference on Information Management (CIKM 2010),

o [HS’05]Haase, P., Stojanovic, L.: Consistent evolution of OWL ontologies. In: ESWC. (2005)

Page 93: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

93

References

o [KPSCG’06] Kalyanpur, A., Parsia, B., Sirin, E., Grau, B.C.: Repairing unsatisfiable concepts in OWL ontologies. In: ESWC. (2006) 170–184

o [Wins’90] Updating Logical Databases. 1990. Cambridge University Press.

o [KM’91] Katsuno H., Mendelzón A. 1991. On the difference between updating a knowledge base and revising it. In Proc. of KR, 387-394.

o [EG’92] On the complexity of propositional knowledge base revision, updates and counterfactuals. In Proc. of AI 57, 227-270.

o [LLMW’06] Liu H., Lutz C., Milicic M., Wolter F. 2006. Updating description logic ABoxes. In Proc. of KR, 46-56.

o [DGLPR’09] De Giacomo G., Lenzerini M., Poggi A., Rosatti R. 2009. On instance-level update and erasure in description logic ontologies. JLC, 745-770.

o [QD’09] Qi G., Du J. 2009. Model-based revision operators for terminologies in description logics. In Proc. of IJCAI, 356-365.

Page 94: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

94

References

o [CKNZ’10] Calvanese D., Kharlamov E., Nutt W., Zheleznyakov D. 2010. Evolution of DL-Lite Knowledge Bases. In Proc. of ISWC, 112-128.

o [WWT’10] Wang Z., Wang K., Topor R.W. 2010. A new approach to knowledge base revision in DL-Lite. In Proc. of AAAI.

o [KZ’11] Kharlamov E., Zheleznyakov D. 2011.Capturing Instance-level ontology evolution for DL-Lite. In Proc. of ISWC, 321-337.

o [CJKZ’12] Cuenca Grau B., Jiménez-Ruiz E., Kharlamov E., Zheleznyakov D. 2012. Ontology evolution under semantic constraints. In Proc. of KR.

o [JRCGHB’11] Jimenez-Ruiz, E., Cuenca Grau, B., Horrocks, I., Berlanga, R.: Supporting concurrent ontology development: Framework, algorithms and tool. DKE. 70:1 (2011)

o [FDCM’08] N. Fridman Noy, T. Tudorache, S. de Coronado, M. A. Musen, Developing biomedical ontologies collaboratively, in: Proc. of AMIA, 2008.

Page 95: Evolution of OWL 2 QL and EL Ontologies Bernardo Cuenca Grau, Ernesto Jiménez-Ruiz Computer Science Department, University of Oxford, UK Evgeny Kharlamov,

95

References

o [MDM’06] A.deMoor,P.DeLeenheer,R.Meersman,DOGMA-MESS:Ameaningevolutionsupport system for interorganizational ontology engineering, in: 14th International Conference on Conceptual Structures, ICCS, vol. 4068 of Lecture Notes in Computer Science, Springer, 2006, pp. 189–202.