Copyright: SIPC From Ontology to Data Model: Choices and Design Decisions Matthew West Reference Data Architecture and Standards Manager Shell International.

Cop

yri

gh

t: S

IPC

From Ontology to Data Model: Choices and Design Decisions

Matthew West

Reference Data Architecture and Standards Manager

Shell International Petroleum Company

Ontolog Forum

“Database and Ontology” mini-series Session-2

16-Nov-2006

2

Introduction

• An ontology describes the types of things that exist, and rules that govern them, a data model defines records about things, and is the basis for a database design. Not all the information in an ontology may be needed (or can even be held) in a data model and there are a number of choices that need to be made. For example, some of the ontology may be held as reference data instead of as entity types. This talk will explore these choices, with illustrations from the ISO 15926-2 data model.

3

Relationship between an Ontology and Enterprise Systems

• Ontology

– Rules

– Classes

– Objects

• Enterprise Systems

– Business Processes (manuals)

– Application Processes and Stored Procedures

– Database Structure

– Records

4

Relationship between an Ontology and Enterprise Systems

• Ontology

– Rules

– Classes

– Objects

• Enterprise Systems

– Business Processes (manuals)

– Application Processes and Stored Procedures

– Database Structure

– Records

5

Relationship between an Ontology and Enterprise Systems

• Ontology

– Rules

– Classes

– Objects

• Enterprise Systems

– Business Processes (manuals)

– Application Processes and Stored Procedures

– Database Structure

– Records

6

Relationship between an Ontology and Enterprise Systems

• Ontology

– Rules

– Classes

– Objects

• Enterprise Systems

– Business Processes (manuals)

– Application Processes and Stored Procedures

– Database Structure

– Records

7

Relationship between an Ontology and Enterprise Systems

• Ontology

– Rules

– Classes

– Individuals

• Enterprise Systems

– Business Processes (manuals)

– Application Processes and Stored Procedures

– Database Structure

– Records

Hard

er (m

ore

expensiv

e) to

ch

ange

Easie

r to e

nfo

rce

limited

8

Data Models: Capabilities and Limitations

thing

X class of X

some X

some other X

Subtype/ supertype relationship

member of

9

Data Models: Capabilities and Limitations

thing

X class of X

some X

some other X

Subtype/ supertype relationship

Member of relationship

My X a member of My class of X

10

Data Models: Capabilities and Limitations

thing

X class of X

some X

some other X

Subtype/ supertype relationship

Member of relationship

My X a member of My class of X

11

Data Models: Some Options

thing

X class of X

Subtype of

member of

12

Data Models: Some options

thing

X class of X

Subtype of

X

some X

some other X

member of

13

Data Models: Some options

thing

class of X

Subtype of

X

some X

some other X

14

Reference Data Model and the Business Information Model level

Product•Name•Price•Price Currency•Date Introduced•Date Withdrawn

Market•Name•Start Date•End Date

Brand•Name•Date Introduced•Date Withdrawn

Sold in Sold under

15

A (3D) Logical Data Model for some Reference Data

Identification•Start Date•End Date

Thing Name• String

Product• Date Introduced

• Date Withdrawn

Market•Start Date•End Date

Brand• Date Introduced

• Date Withdrawn

Selling of productunder Brand•Start Date•End Date

Selling of product in Market•Start Date•End Date

of identifier

in of

ofunder

Product Price• Value•Start Date•End Date

Currency•Start Date•End Date

in

for

16

A denormalized Physical Data Model for Product, Brand, and Market

PRODUCT_T• ID (Key)• Version (key)• Name• Brand1ID (FK)• Brand1Name• Brand2ID (FK)• Brand2Name• Market1ID (FK)• Market1Name• Market2ID (FK)• Market2Name• Price• PriceCurrency

(FK)• DateCreated• DateLastModified• DateWithdrawn

BRAND_T• ID (Key)• Version (Key)• Name• DateCreated• DateLastModified• DateWithdrawn

MARKET_T• ID (Key)• Version (key)• Name• DateCreated• DateLastModified• DateWithdrawn

17

A Generic Physical Data Model for Reference Data

THING_T•ID•ThingType (FK)•StartDate•EndDate

ASSOCIATION_T•ID•AssociationType (FK)•Table1•ID1 (FK)•Table2•ID2 (FK)•StartDate•EndDate

VALUE_T•ID•PropertyType (FK)•ValueType (FK)•String•Number•UnitOfMeasure

18

Some Data Modelling Choices made in ISO 15926-2

• Relations modelled as entity types (not attributes)

• Key entity types for 4D paradigm included

• Subtypes of X not held when they can be held as instances of class_of_X, especially when numbers of subtypes are large

• Explicit model of representation (names and numbers) included to support multiple names and descriptions

• Meta-model included for extensibility, the ontology makes an open world rather than closed world assumption

19

Some choices made in ISO 15926

(ABS)thing

16,1 possible_individual

6,1 possible_individual

(ABS)abstract_object

2,1 class

18,2 representation_of_Gregorian_date_and_UTC_time

18,2 representation_of_Gregorian_date_and_UTC_time

18,2 representation_of_Gregorian_date_and_UTC_time

17,1 class_of_information_representation

11,1 relationship

4,1 multidimensional_object

*id STRING

record_copy_created

record_created

record_creator

record_logically_deleted

why_deleted

20

Some choices made in ISO 15926

possible_individual

arranged_individual

9,1 event

period_in_time

physical_object

materialized_physical_object

functional_physical_object

stream

spatial_location

9,2 activity

actual_individual

whole_life_individual

composition_of_individual

1

arrangement_of_individual

1

assembly_of_individual

feature_whole_part

(RT) whole

whole

temporal_whole_part

9,3 participation

9,4 temporal_bounding

part

Note: has no subtypes

21

Some choices made in ISO 15926

class_of_arranged_individual

1

class_of_atom

phase

class_of_biological_matter

class_of_functional_object

class_of_compound

class_of_molecule

class_of_composite_material

crystalline_structure

class_of_sub_atomic_particle

class_of_information_presentation

17,1 class_of_information_representation

class_of_particulate_material

class_of_organization

10,1 class_of_activity

class_of_information_object

class_of_feature

1

class_of_organismclass_of_person

class_of_inanimate_

physical_object

Subtypes of arranged_individual can be found here as instances

22

The split of classes between entity types and instances in ISO 15926

ISO 15926-2

Generic Classes

Standard Classes

Proprietary Classes

Data Model (201 classes)

Reference Data Library Standard and Proprietary classes (some 24,000 classes currently under development)

Reference Data Library generic classes (some 10,000 classes due for publication as ISO 15926-4)

23

Representation (by e.g. text or numbers)

6,1 possible_individual

6,1 possible_individual

6,1 possible_individual

representation_of_thing

definition(RT) represented

represented 1,1 thing

2,1 class

identification

description

sign

usage_of_representation

used user

responsibility_for_

representation

controlled controller

e.g. piece of paper with text on it.

24

Class of representation

6,1 possible_individual

6,1 possible_individual

class_of_information_representation

1

18,1 class_of_EXPRESS_information_representation

18,2 representation_of_Gregorian_date_and_UTC_time

class_of_representation_

translation

class_of_first

class_of_second

class_of_usage_of_representation

class_of_used

class_of_representation_

of_thing

class_of_identification

class_of_definition

(RT) represented

represented 1,1 thing

2,1 class

class_of_description

pattern

user

class_of_responsibility_for_

representation

class_of_controlled controller

e.g. the text on one or more pieces of paper.

25

Part of the Meta-Model

class

specialization

subclass

superclass

classificationclassified

1,1 thingclassifier

26

Another part of the Meta-Model

role_and_domain

role

class_of_relationship_

with_signature

specialization_by_domain

(RT) subclass

specialization_by_role

(RT) subclass

(RT) superclass

class_of_end_1

class_of_end_2

27

From Data Model to Database - Considerations

• First cut, each entity type maps to a table, but only if

– Each table has some distinct attributes, or

– There is some reason to segregate records (volume or access requirements)

28

From Data Model to Database

possible_individual

arranged_individual

9,1 event

period_in_time

physical_object

materialized_physical_object

functional_physical_object

stream

spatial_location

9,2 activity

actual_individual

whole_life_individual

29

From Data Model to Database

possible_individual

arranged_individual

9,1 event

period_in_time

physical_object

materialized_physical_object

functional_physical_object

stream

spatial_location

9,2 activity

actual_individual

whole_life_individual

PossibleIndividuals

•IsWholeLifeIndividual

•IsActualIndividual

•IsArrangedIndividual

•IsEvent

•IsPeriodInTime

•IsActivity

•IsPhysicalObject

•IsFunctionalPhysicalMaterial

•IsStream

•IsSpatialLocation

30

Conclusions

• Data Models are a key component for getting ontologies used in Information Systems

• There are a wide range of options when making design decisions both in moving from Ontology to Data Model/Reference Data, and from Data Model to Database

• Which options are best is likely to depend on the particular circumstances of and purposes of an application