ontological analysis and conceptual modelling: achievements and perspectives

Post on 05-Jul-2015

351 Views

Category:

Presentations & Public Speaking

0 Downloads

Preview:

Click to see full reader

DESCRIPTION

FOIS 2014 keynote

TRANSCRIPT

1

Ontological Analysis !and Conceptual Modelling: !

Achievements and PerspectivesNicola Guarino!

!National Research Council of Italy!

Institute of Cognitive Sciences and Technology!Laboratory for Applied Ontology, Trento!

!(Ongoing work with Giancarlo Guizzardi, Federal University of Espirito Santo, Brazil)!

1

Summary

2

1. Knowledge representation, conceptual modelling, and ontological analysis!

2. The ontological level, 20 years later!

3. OntoClean, 13 years later!

4. Roles and relationships, 23 years later!

5. Dolce, 11 years later!

6. Episode-driven conceptual modeling!

7. A future challenge for ontological analysis

DisclaimerUnfortunately, this is not a comprehensive state of the art.

Just some subjective considerations mainly from the perspective of my own present interests.!

I wish I could have done more :-(

• Ontological analysis: systematic way to understand and make explicit the world assumptions behind a certain description:!• How do we believe the world is, when we say !

• This rose is red!• John is married with Mary!• John is a student!• My name is Nicola

• Data models encode knowledge about the world in order to easily and efficiently access it!

• Conceptual models describe some aspects of the world for the purpose of understanding and communication!

• Knowledge representations encode knowledge about the world in order to easily and efficiently access it and use it to generate new knowledge!

• Computational ontologies characterize the language used to talk about the world in order to reduce ambiguities and misunderstandings

Knowledge representation, conceptual modeling, and ontological analysis

3

different role of axioms

possible expansions of domain !and vocabulary

Ontological analysis as a detective lens

• Most of our true statements about the world are approximate!

• What makes them true?!• Where…?!• When…?!• Who…?!

• Why…?!

• Ontological analysis as a search for Truth-makers

4

The Ontological Level, 20 Years Later

5

The Ontological Level

Level Primitives Interpretation Main feature

Logical Predicates, functions

Arbitrary Formalization

Epistemological Structuring relations

Arbitrary Structure

Ontological Ontological relations

Constrained (meaning postulate s )

Meaning

Conceptual Conceptual relations

Subjective Conceptualization

Linguistic Linguistic terms

Subjective Language dependence

• Guarino N. 1994. The Ontological Level. In R. Casati, B. Smith and G. White (eds.), Philosophy and the Cognitive Sciences (by 16th International Wittgenstein Symposium, Kirchberg am Wechsel, Austria, 1993). Vienna, Hölder-Pichler-Tempsky 1994!

• Guarino, N. 2009. The Ontological Level: Revisiting 30 Years of Knowledge Representation. In Alex Borgida, Vinay Chaudhri, Paolo Giorgini, Eric Yu (eds.), Conceptual Modelling: Foundations and Applications. Essays in Honor of John Mylopoulos, Springer Verlag 2009

From the logical level to the ontological level

• Logical level (no structure, no constrained meaning)!• ∃x (Apple(x) ∧ Red(x))!

!• Epistemological level (structure, no constrained meaning):!

• ∃x:apple Red(x) (many-sorted logics)!• ∃x:red Apple(x)!• a is a Apple with Color=red (description logics)!• a is a Red with Shape=apple!

!• Ontological level (structure, constrained meaning)!

• Some structuring choices are excluded because of ontological constraints: Apple carries an identity condition, Red does not.!!

Ontology helps building “meaningful” representations

OntoClean, 13 years later!(joint work with Chris Welty, Vassar College, !

then at IBM Watson Research Center, now at Google Inc.)

8

Ontoclean 13 years later • Basic idea: !

• assumptions about individuation and identity constrain taxonomic relationships: is-a overloading can be controlled!

• such assumptions are a basic component of ontological commitment!• Large impact, despite:!

• difficulties in coming up with identity precise criteria (necessary and sufficient conditions)!

• some problems in the definitions for rigidity and identity criteria (discussed and fixed in various papers by Carrara&Giaretta, Welty…)!

• Implementations in Protege and in other software modelling tools !• Some tools for deciding identity conditions on the basis of NL questions

have been developed [Oltramari…]!• Simplifications: !

• just focus on identity conditions (either only necessary or only sufficient)!

• unity criteria (a kind of weak identity criterion) are enough in many cases

9

10

From OntoClean meta-properties to a basic ontology of properties (universals)

Property

Non-sortal -I

Role ~R+D

Sortal+I

Formal Role

Attribution -R-D

Category +R

Mixin -D

Type +O

Quasi-type -O

Non-rigid -R

Rigid +R

Material roleAnti-rigid

~R Phased sortal -D

UFO and OntoUML !(Giancarlo Guizzardi and coll., UFES, Brazil)

• UFO: A foundational ontology that shares many of the basic intuitions of DOLCE, and includes an ontology of universal largely based on OntoClean!

• OntoUML: A visual conceptual modeling language whose primitives reflect ontological distinctions put forth by UFO!

• Besides the OntoUML language itself, the approach includes a number of methodological principles, design patterns and computational tools for doing formal verification, validation (via visual simulation), verbalization and automated transformation to operational languages such as OWL!

• Adopted in many industrial projects in different domains (telecommunications, government, digital journalism, software engineering, etc.)

11

Parts and wholes, 18 years later• 1996 paper!• Parts and moments!• DKE special issue!• Functional parts!• Debate at the summer school!• mereogeometry!• Unity and plurality

12

Roles and relationships, 23 years later

13

Constraining structuring relationships

Woods’ “What’s in a link?” (1975): JOHN HEIGHT: 6 FEET KISSED: MARY !

"no longer do the link names stand for attributes of a node, but rather arbitrary relations between the node and other nodes”

14

Structuring relations: a broader picture

JOHN HEIGHT: 6 FEET RIGHT-LEG: LEG#1 MOTHER: JANE JOB: RESEARCHER KISSED: MARY

intrinsic quality part role

relational quality external relation

We need different primitives to express different structuring relationships among concepts We need to represent non-structuring relationships separately Current description logics tend to collapse EVERYTHING!

15

structuring relationships

Roles, a never-ending story• LOT of literature, but still several open issues!

• AO special issue on roles (Boella)!• KR 2004 paper on social roles (Masolo et al.)!• Extensive further work by Mizoguchi, Loebe, Kassel….!

• In OntoClean we defined roles as anti-rigid, dependent properties!• Problems: broken, widow, driving…!

• For some, roles are particulars, not (reified) properties!• An old debate: “Roles are not classes: a reply to Nicola

Guarino” (van Heijst et al. 1997)!• Going back to their etymology: roles as descriptions (abstract

particulars)?!• Maybe we can take descriptions for the proper referent of the term

‘role’…!• …and distinguish between roles and the corresponding properties

(role properties)!• …but there are still problems

16

Being in a role vs. playing a role!some striking linguistic behaviours of social roles

17

• Comparatives:!• John has a good friend!• *John has a good classmate!• He is a good driver !• *he is a good pedestrian/passenger!• He is a good president !• *he is a good ocular witness!• *she is a good widow!!

• Replaceability:!• John replaced his teacher!• *John replaced his friend!• The company replaced an employee!• *The company replaced a customer

• Playing a social role does not mean instantiating the properties expected to be associated to the role!

• Some social roles (but not all!) are “replaceable”.

US president

Nixon

Quaker

embodies The US Presidency

US Presidency

satisfies

Natural personLegal person

~R +R~R

Nixon’s diamond revisited

Us president role

plays

Quaker role

plays

describes?describes?

variable embodiments emerge only for assigned (functional) roles

Dolce, 11 years later

19

Dolce, 11 years later

• Very robust and stable top ontology!• Several applications, various core or domain ontologies based on it!• Used for cleaning up and improving lexical resources [Oltramari, Vieu…]!• Just one paper discussing its foundational choices [Borgo&Masolo]!• Individual qualities as a crucial notion!• Three needed extensions, in my opinion (among many others!):!

• Local qualities!• Relative qualities!• A more detailed account of perdurants and the different ways of

participation/involvement

20

The four-category ontology

21

Kinds Attributes

Objects Modes

instantiated byinstantiated by

characterised by

characterised by

Substantial universals

Substantial particulars Non-substantial particulars

Non-substantial universals

Dolce qualities and the four-category ontology

22

KindsQuality kinds

Objects Qualities

instantiated byinstantiated by

characterised by

characterised by

Substantial universals

Substantial particulars Non-substantial particulars

Non-substantial universals

Quality manifesta

tions

instantiated by

Qualia

manifest as

Quality spacescharacterised by

atomic parts of

Dolce and the four-category ontology!(a rough attempt)

23

Kinds

Objects Qualities

instantiated by

characterised by

characterised by

Substantial universals

Substantial particulars Non-substantial particulars

Non-substantial universals

Events

instantiated by

constitutive subjectsof

Properties

exemplified by

Events (perdurants) and their participants

• Lombard: an event is a change of an object in a quality space!• The simplest event I can think of is the manifestation of an individual

quality: !• the redness of this rose lasted one week!

• Focusing on the ultimate constituent subjects of events (i.e., individual qualities) allows us to distinguish events involving the same objects at the same time!

• Isn’t an event, after all, what happens to a sum of individual qualities?!

• Giving an event, focusing on its constituent qualities helps recognising it!

• Note: some of these individual qualities can be relational qualities (e.g., the distance of the Earth from the Sun) or local qualities (e.g., the depth of Adriatic Sea near Ravenna)

24

What's the width of a river?

25

What's the depth of the Adriatic Sea?

26

What’s the width of a vase?

What’s the color of a car?

Referring to local qualities

27

• Some qualities inhering to parts are conceived as if they where inhering to the whole:!? the width of this river stretch is 100 meters!√ the river’s width is 100 meters here!

? the Croatian side of the Adriatic Sea is much deeper than the Italian side !√ the depth of the Adriatic Sea is much higher along the Croatian coast than along the Italian coast

The problems with local qualities

...Of course, this only happens for some quality kinds:!the river has (arguably) just one length (at a given time)!the car has just one mass (at a given time)!

• Problems!• What is a local quality?!• What are the quality kinds that “generate” local qualities? !• In which sense does a local quality, which inheres to a

proper part, also “inhere” to the whole?!• How to make sense of terms like the color of the car even

if the actual color varies a lot?

28

Local to what?• Is the width of a longitudinal part of the river a

width of the river?

29

• Is the width of the handle a width of the vase?!• How can we exclude this?!• Answer: conventional parts

Episode-centric conceptual modeling

30

Objects and Episodes• DOLCE distinguishes between entities that endure (persist) in time and

entities that perdure (happen) in time: a person is an endurant, a talk is a perdurant!

• Ordinary endurants are commonly called objects. !• No standard term is used for ordinary perdurants:!

• events, happenings, circumstances, occasions, situations, processes, (facts, deeds)…!

• Episode is the generic term I propose for a large class of relevant perdurants.!• “An episode is an event, a situation, or a period of time that is important or

interesting in some way” (Oxford Advanced Learning Dictionary)!• Why “Episode” may be a good choice for conceptual modeling:!

• Focuses on something relevant which is happening!• Conveys a unity criterion (maximality)

31

Episodes as maximal occurrents

32

jumping episode

jumping process

jumping process

Episodes as truth makers• Kinds of properties!

• permanent:!• essential: Person(John), Number(3)!• contingent: !

–intrinsic: LongNosed(John)!–extrinsic: Brazilian(John)!

• temporary:!–intrinsic: Tired(John)!–extrinsic: InLoveWithMary(John)!

!• What are their truth makers?!

• Person(John) holds because John exists. John is the truth maker of the proposition. Nothing else is necessary.!

• Tired(John), in addition to the existence of John, requires an episode to occur at a certain time. It is such episode (involving John) that is the truth-maker.!

• All temporary properties (both intrinsic and extrinsic) require an episode as their truth maker.!

• Some (all?) permanent extrinsic properties require an episode as their truth maker.33

Relations and their truth makers

• Kinds of relations (a rough taxonomy - More work to be done!)!• Permanent relations!

• Essential: greater(3,2) !• Contingent: !

• intrinsic: same-blood-group(John, Mary)!• extrinsic: born-in(John, Brazil)!

• Temporary relations!• Intrinsic: taller(John, Mary)!• Extrinsic: loves(John, Mary)!!

• What is their truth maker?!• All temporary relations require an episode as their truth maker!• Some permanent relations (the extrinsic ones) require an episode

as their truth maker

34

Episode-centric conceptual modeling:!put truth-makers explicitly in your domain

• Why is it so useful?

35

In an episode there is more!...

• Whenever a contingent relation holds, you can ask yourself the usual questions:!

• When?!• Where?!• Who?!• What?!• Why?!

!• …and include the corresponding knowledge in your model only if you put

episodes in your domain!!• ….That’s why we can’t confuse episodes with facts or reified relations

The expressive power of episodes:!revisiting Guizzardi’s notion of ‘relator’ (*)

36

(*) Giancarlo Guizzardi, Ontological Foundations for Structural Conceptual Models, University of Twente, 2005

Episodes  solve  the  cardinality  constraint  problem

«role»Patient

«kind»Medical Unit

1..*1..* treated In

A material relation:

Episodes,  Relators  and  Derived  Material  Relations

«role»Patient

«kind»Medical Unit

«relator»Treatment

1..*

1

«mediation»

1

«mediation»

«kind»Person

1..*

1..* 1..*

/treated in

1

1..*

‹‹ material  ››

Indeed this is an episode

has-patient has-agent

A closer look at relationships

• Chen 76: “a relationship is an association among entities. [However,] some people may view a marriage as an entity while other people may view it as a relationship. We think that this is a decision which should be made by the Enterprise Administrator”!

• Rationale: relationships can bear properties: a project-worker relationship can have the attribute percentage-of-time representing an intrinsic property of the relationship itself.!

• Such attributes are actually properties of episodes…!• Database people distinguish between relation constraints and tuple constraints

(such as those involving the number of patients being treated by a medical unit at the same time)!

• Tuple constraints do not actually constrain relations, but their truth-makers (episodes) !

• If so, what is a relationship?!• a tuple?!• an episode?!• a bundle of relational qualities?

39

From formal ontology to applied ontology!a glimpse the future

40

From ontology-driven information systems to ontology-driven socio-technical systems

top related