Textual and Logical Definitions in Ontologies Selja Seppl

  • Slides: 33
Download presentation
Textual and Logical Definitions in Ontologies Selja Seppälä, Yonatan Schreiber and Alan Ruttenberg October

Textual and Logical Definitions in Ontologies Selja Seppälä, Yonatan Schreiber and Alan Ruttenberg October 7, 2014 IWOOD 2014 Houston, USA

BACKGROUND IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg

BACKGROUND IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 2

Issue • Ontologies have axioms and natural language definitions • But no theory of

Issue • Ontologies have axioms and natural language definitions • But no theory of their respective functions Varying authoring practices No specific recommendations or guidelines IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 3

Observation • There are correspondences between the logical and textual parts, both structurally and

Observation • There are correspondences between the logical and textual parts, both structurally and functionally Recommendation • There should be correspondence between the contents of the textual and logical definitions as well IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 4

Proposal • Exploit these correspondences to – Build tools for definition and axiom checking

Proposal • Exploit these correspondences to – Build tools for definition and axiom checking – Provide guidelines for improving textual definitions and axioms • Bring automated methods used in terminology to – Establish the correspondences – Identifying quality issues in textual definitions – Map to quality issues in the axioms IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 5

Outline • Textual definitions from the terminologist’s perspective • Axioms in ontologies from the

Outline • Textual definitions from the terminologist’s perspective • Axioms in ontologies from the logician’s perspective • Correspondences between textual and logical definitions • Leveraging the correspondences to check definitions • Conclusion IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 6

TEXTUAL DEFINITIONS IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A.

TEXTUAL DEFINITIONS IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 7

A Terminological Account • Concise sentence • Object of an annotation property • Also

A Terminological Account • Concise sentence • Object of an annotation property • Also found in terminological dictionaries Account based on (Seppälä 2012, forthcoming) IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 8

A Good Definition • Conveys the intended meaning of an ontology term • Describes

A Good Definition • Conveys the intended meaning of an ontology term • Describes the type of thing to which the term refers by – Genus: states the type of thing to which the instances of the defined term belong – Differentia: distinguishes these instances from the type and from other things falling under the same type by listing one or more of the characteristics of the instances of the term A definition has a structure IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 9

Example from the Cell Type Ontology (CL) genus: type of thing leukocyte An achromatic

Example from the Cell Type Ontology (CL) genus: type of thing leukocyte An achromatic cell of the myeloid or lymphoid lineages capable of ameboid movement, found in blood or other tissue. differentiae: distinguishing characteristics IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 10

The Structure of a Definition • Each part is related to the defined term’s

The Structure of a Definition • Each part is related to the defined term’s instances by some type of relation – Genus: is_a – Differentia: any kind of relation IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 11

Relations Describing the Structure is_a develops_from leukocyte An achromatic cell of the myeloid or

Relations Describing the Structure is_a develops_from leukocyte An achromatic cell of the myeloid or lymphoid lineages capable of ameboid movement, found in blood or other tissue. capable_of IWOOD 2014 | October 7, 2014 located_in S. Seppälä, Y. Schreiber and A. Ruttenberg 12

Functions of Textual Definitions • Allow users to make their use of a term

Functions of Textual Definitions • Allow users to make their use of a term converge toward that of the rest of the users of the ontology by – Specifying terms’ intended meaning – Disambiguating terms with necessary and sufficient conditions ontologies contain only classical definitions • Give non-ontologists more detailed and explicit information about the term and its referent IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 13

AXIOMS IN ONTOLOGIES IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and

AXIOMS IN ONTOLOGIES IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 14

What Are Axioms? • Expressions in a formal language (in this paper we consider

What Are Axioms? • Expressions in a formal language (in this paper we consider primarily OWL) restricting the use of a term • Axioms in OWL impose necessary conditions on the use of a term • It is rarely possible to provide conditions that are jointly sufficient • Axioms thus do not, on their own, constitute complete definitions IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 15

Axioms and Term Usage • Disambiguation and OWL Reasoners: – The reasoner can check

Axioms and Term Usage • Disambiguation and OWL Reasoners: – The reasoner can check a dataset for consistency with an ontology – If assertions are made in the data that violate restrictions imposed by the axioms, the dataset will be found inconsistent – Axioms thus restrict the use of terms to design intention IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 16

Axioms and Term Usage • Axioms provide models for RDF expressions SSR_Abstract IAO: is_about

Axioms and Term Usage • Axioms provide models for RDF expressions SSR_Abstract IAO: is_about Seppala. Schreiber. Ruttenberg 2014 SSR_Abstract ro: part_of Seppala. Schreiber. Ruttenberg 2014 IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 17

Functions of Axioms • Axioms, like textual definitions, help to regularize the use of

Functions of Axioms • Axioms, like textual definitions, help to regularize the use of terms among users – Logically restrict the use of terms by means of the reasoner – Provide schemata for common assertions IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 18

CORRESPONDENCES BETWEEN TEXTUAL AND LOGICAL DEFINITIONS IWOOD 2014 | October 7, 2014 S. Seppälä,

CORRESPONDENCES BETWEEN TEXTUAL AND LOGICAL DEFINITIONS IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 19

Recommendation Textual definitions and axioms should have the same content, as far as possible.

Recommendation Textual definitions and axioms should have the same content, as far as possible. IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 20

Failures of Correspondence 1. Incomplete textual definitions 2. Missing axioms, or axioms with less

Failures of Correspondence 1. Incomplete textual definitions 2. Missing axioms, or axioms with less precise content than the textual definitions 3. Redundant content of axioms or definitions IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 21

1. Incomplete Textual Definition IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber

1. Incomplete Textual Definition IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 22

2. Missing or Imprecise Axioms IWOOD 2014 | October 7, 2014 S. Seppälä, Y.

2. Missing or Imprecise Axioms IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 23

3. Redundant Content IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and

3. Redundant Content IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 24

LEVERAGING THE CORRESPONDENCE TO CHECK DEFINITIONS IWOOD 2014 | October 7, 2014 S. Seppälä,

LEVERAGING THE CORRESPONDENCE TO CHECK DEFINITIONS IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 25

Semi-Automated Definition Checking • The correspondence we described can be used to automatically compare

Semi-Automated Definition Checking • The correspondence we described can be used to automatically compare textual definitions with axioms for every term in an ontology • This will not automatically fix mismatches, but it can flag them and indicate the nature of the mismatch IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 26

Semi-Automated Definition Checking 1. Do any of the terms from the ontology appear in

Semi-Automated Definition Checking 1. Do any of the terms from the ontology appear in the textual definition? 2. Get the list of ancestor classes for any such terms. 3. Do any of these classes appear among the relata in the axioms? 4. If no correspondence between terms in textual definition and classes in the axioms do any of the relations expressed in the differentiae of the textual definition correspond to objectproperties in the axioms? 5. Tag the corresponding parts of the textual definitions with the matching ontology terms (this could provide links in the textual definition to the relevant classes in the ontology). 6. If mismatches manually correct, modify, or complete either the textual definition, the axioms or both. IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 27

CONCLUSION IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg

CONCLUSION IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 28

 • Textual definitions and axioms serve complementary functions in ontologies: in both cases

• Textual definitions and axioms serve complementary functions in ontologies: in both cases their primary functions are disambiguation, and regularization of term usage • Since it is important for the textual definitions and the axioms of an ontology to determine the same usage, their content should be parallel wherever possible • Because of the genus/differentia structure of textual definitions, it is often possible to obtain a close correspondence with the axioms • We have proposed a semi-automated system for checking the correspondences between the textual definitions and the axioms in an ontology IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 29

ADDITIONAL SLIDES IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A.

ADDITIONAL SLIDES IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 30

Logical forms of textual definitions • Derive from the relationship between their intension and

Logical forms of textual definitions • Derive from the relationship between their intension and their extension • Classical definition: the intension holds for all instances of the defined type – triangle: A rectilinear figure that has three sides. • Typical or prototypical definition: the intension holds for most of the instances – swan: An aquatic bird with a long neck, usually having white plumage. • Instance definition: the intension holds for only a single instance – Large Hadron Collider (LHC): The world’s largest and most powerful particle accelerator. IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 31

Axioms for Taxonomy Checking IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber

Axioms for Taxonomy Checking IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 32

Checking textual definitions • Automatically loook for – Particularizing expressions (for example, especially, in

Checking textual definitions • Automatically loook for – Particularizing expressions (for example, especially, in particular, i. e. , etc. ) & punctuation (parentheses, colons) – Overly generalizing expressions (etc. , in general, normally, etc. ) Use other annotation properties for non-definition contents – Definitions of other terms: • Punctuation (parentheses, colons) • Expressions introducing new information (i. e. , that is) IWOOD 2014 | October 7, 2014 S. Seppälä, Y. Schreiber and A. Ruttenberg 33