Mapping provenance in ontologies

Transcription

1 Mapping provenance in ontologies František Jahoda FNSPE Czech Technical University Institute of Computer Science Academy of Sciences of the Czech Republic

2 What is Provenance? from the French provenir, "to come from" confirm the time, place, and if appropriate the person responsible, for the creation, production or discovery of the object. artefact's complete documented history no clear semantics... Mapping provenance in ontologies 2

3 Provenance Is used in Archaeology, Paleontology Arts, Archives, Science Is used for to discover a forgery, looted, reproduction to prove ownership can set the price of antiques to prevent evidence tampering Chain of custody To prove ecology behaviour (wood, endangered species) Mapping provenance in ontologies 3

4 Data provenance Is used for Managing trustworthiness of data sources Provide provability of conclusions Recording history (not only actual snapshot) Enables learning from history, sharing data (data rights management) To prevent conflicts between data from two different sources Mapping provenance in ontologies 4

5 Ontology provenance Ontologies have different life cycle Less control over ontology sources Ontologies are more rigid than data Part of the ontology creates bridge between abstract model and concrete data Logical nature of ontologies have to be considered Derived axioms Change of one axiom can change meaning of others Mapping provenance in ontologies 5

6 Software project life cycle (Waterfall) Requirements Design Implementation Verification Maintenance Mapping provenance in ontologies 6

7 Ontology life cycle Create/Select Development and/or Selection Evolve Extension/Modification Populate Knowledge base generation Validate Consistency checks Maintain Usability tests Deploy Knowledge retrieval Mapping provenance in ontologies 7

8 Changes in ontology Understanding of subject area is developing e.g. three fish species can be only one species in different age Input data have changed meaning An error in ontology was found New requirements / extending ontology Speed up Mapping provenance in ontologies 8

9 How to implement provenance? Encoding information about ontology Information about ontology should have own ontology. (answering queries) Example of provenance data model How to encode derived axioms? Mapping provenance in ontologies 9

10 Encoding information about ontology in OWL DL Statements about axioms, higher order statements are not possible Annotations compatible with OWL 1.0 Reifying axioms Annotation properties Annotations using axioms URI Mapping provenance in ontologies 10

11 Axioms reifying Axiom in original ontology PERSON MORTAL Using the meta ontology, we can represent this ontology as follows: CLASS(Mortal) CLASS(Person) SUBCLASS_OF_AXIOM(axiom1) SUBCLASS_OF_SUPERCLASS (axiom1, Mortal) SUBCLASS_OF_SUBCLASS (axiom1, Person) SUBCLASS_OF (Person, Mortal) Mapping provenance in ontologies 11

12 Axioms reifying Adding annotations CREATOR (axiom1, Aristotle) CONFIDENCE (axiom1, 0.95) But meta ontology is considerably bigger.. Mapping provenance in ontologies 12

13 Annotation properties Using Annotation properties (same ontology) Example PERSON MORTAL (axiom1) CREATOR (axiom1, Aristotle) CONFIDENCE (axiom1, 0.95) But, provenance is treated outside the OWL logical semantics Mapping provenance in ontologies 13

14 Annotation with metaontology Using Meta Ontology (separate ontology) Extension in OWL 2.0 draft Example (in original ontology) PERSON MORTAL (axiom1) In meta ontology AXIOM(axiom1) CREATOR (axiom1, Aristotle) CONFIDENCE (axiom1, 0.95) Mapping provenance in ontologies 14

15 Example of provenance ontology Where Location Why Decision Rationale S 1:1 S Position S Time -varying Position occurs in because of 1:M 1:M Goal Belief Point Line Region 0:M 1:1 leads to Assumption S Hypothesis Event When Valid Time S 1:1 happens at Instant Time Period Time Interval 0:M 0:M is involved in 0:1 causes 1:1 How method precondition artifact Who Role Agent 1:M 1:M performs 1:m 0:M Action S resourc e Position S Individual Organization Artificial Agent Group action dependency Primitive Action Complex Action Mapping provenance in ontologies 15 Group

16 Example of provenance ontology Representation of history as a sequence of events which cause changes of state. Events have properties: When event time (point, interval) Temporal granularity (day, business day, hour) Where place of event (point, line, region) Position can change with time Mapping provenance in ontologies 16

17 Example of provenance ontology How actions leading to the event Primitive actions, complex actions Preconditions, Resources Methods what was done, capture action's parameters Artefacts data manipulated by the action (input/output) Who refers to people and and organisation involved Agents individual, organization, software agent Role functional responsibility (e.g. in organization) Why rationale of an action Goals, Beliefs Assumptions, Hypothesis Mapping provenance in ontologies 17

18 How to encode derived axioms? Each inferred axiom is based on a minimal set of base axioms. Let is union of all axioms in such sets called O AJ() Each base axiom is connected in meta ontology with provenance atoms (describing provenance information, N:M binding). Therefore, axiom should be connected with all provenance atoms related to O AJ() Mapping provenance in ontologies 18

19 References Denny Vrandečić, Johanna Völker, Peter Haase, Thanh Tran Duc, Philipp Cimiano: A Metamodel for Annotations of Ontology Elements in OWL DL Dr. Sudha Ram: the Active Conceptual Modeling of Learning Workshop, Space and Naval Warfare Systems Center, San Diego, May 9 12, 2006 Paul Buitelaar: elearning4nlp Workshop Prague, June 2007 Miroslav Vacura, Vojtěch Svátek: Pattern Based Representation and Propagation of Provenance Metadata in Ontologies, Poster Proceedings EKAW2008 Mapping provenance in ontologies 19