Joost Breuker

1
Ontology, ontologies and ontological reasoning2
what do ontologies represent?

• Joost Breuker
• Leibniz Center for Law
• University of Amsterdam

2
Overview

• Reference to conceptualizations
• Knowledge and semantics
• Kinds of conceptualizations
• Representing/expressing conceptualizations
• LKIF-Core
• Towards a top ontology based on common sense
• A cognitive science perspective
• Main categories explained

3
Conceptualizations and representations

• ontology the specification of a
  conceptualization
• Confusion about what ontologies are vs what they
  represent
• They are representations, i.e. expressions in
  some medium.
• They represent conceptualizations of terms used
  in a domain
• Barry Smiths (2008) illustrates the confusion

4

• Cognitive representations
• Representational artifacts
• Reality

5
Ontologies are here
6
Ontologies are here? But where is the
conceptualization?
7
or here
8
or here? But is this the conceptualization of a
term or of an image??
9
Ontologies do not represent concepts in peoples
heads
10
Ontologies do not represent concepts in peoples
headsbut not this way
11
Like the scientific theories from which they
derive, they represent universals in realitye.g.
leg
12
Like the scientific theories from which they
derive, they represent universals in realitye.g.
leg
13
Cleaning up?

• The scientific theories are conceptualizations
• The leg is an instance of leg
• There are scientific (eg in anatomy) and common
  sense legs (eg nice legs),
• Wheres HERE the ontology?
• Extra misleading due to using 2analog
  representations
• Screen
• Drawing (via some conceptualization
  perception)
• An ontology is a symbolic representation of
  concept (eg leg)
• This representation should capture the meaning

14
Having this clarified, there remain new
ambiguities

• What kind of conceptualizations do ontologies
  capture?
• How we directly understand and communicate the
  world (common sense), or
• Models of worlds derived from controlled
  observation (scientific)
• (Models from speculation (eg meta-physics))
• Ontologies represent
• (Parts of) knowledge (bases, specifications) (AI)
• Repositories of semantics (Semantic Web,
  Computational linguistics)

15
Semantics and knowledge

• What we know about terms vs what terms mean in a
  particular context (domain, document, phrase,..)
• Semantics is the result of applying knowledge to
  data
• It gives meaning to data (signs) ? information
• In the process of understanding, those
  properties of terms are selected that make up a
  coherent macro-structure (model).
• It is contextualized knowledge

16
knowledge and semantics (cf Levelt, Speaking,
CUP 1993, fig 3.1)
semantics (meaning/sense)
semantic representations (preverbal messages)
FORMULATOR
ontologies?
knowledge
17
sense vs meaning
what is this?
this is a car

• in traffic a car is a vehicle, moves,
  transports,

• for the mechanic a car is a device, has a
  motor, etc

• for a car salesman a car is a commodity, has a
  price, a colour, accessories,

• for an insurance inspector is this a car or a
  wreck?

18
context dependency of meaning

• the more abstract, the less properties, and the
  less possible variation in meaning/sense
• context dependency views select properties

19
levels of ontologies

• top, upper, foundational ontologies
• the primitives on which we build our knowledge
  (eg space, time, object, process, substance,
  etc.)
• core ontologies
• some field of practice, discipline (e.g.
  medicine, law, etc.)
• domain ontologies
• the domain of interest, e.g. (Dutch) traffic
  law,

20

• The more abstract, the more meaning overlaps with
  conceptual knowledge
• Domain ontologies are usually capturing
  semantics this limits reuse
• Lingua Universalis view
• Words lt -- gt Concepts
• Now
• Words lt -- gt Meaning lt -- gt Concepts lt -- gt
  Knowledge
• Context

21
What kind of conceptualization?

• Domain ontologies consist usually of a
  specialized terminology and a particular view on
  some world.
• Core ontologies mediate between some top
  ontology and the main concepts in a domain
• Eg medicin ? scientific notions about biological
  processes
• Eg law ? refined terms to interpret common
  sense events (sense)
• Current top ontologies do not take common sense
  seriously, neither do they commit to real
  scientific concepts
• For example space time

22
Space in top ontologies a scientific
perspective

• SUMO, DOLCE, CyC, BFO, etc
• 3 dimensions 3D ontology for objects
  (continuants)
• Time added4D ontology for processes (occurrents)

23
Space is viewed as in outer space (mechanics)

• The 3 dimensions are of equal importance, even
  translated in common-sense terms
• Front-back
• Up-down
• Left-right

24
Looking at space on earth
25
Abstracted
26
Space in a common-sense ontology

• Space for positions/places
• A horizontal plane on which objects can stand
• 2 eyes horizontal front-back ? depth
• Left-right confusion (look at the mirror)
• A vertical axis up-down
• Supported by gravity
• Space of objects (extension)
• 3 D
• Combination
• We can see how
• Objects can survive in constructions or
  assemblages

27
Combining the scientific with the common sense
view
28
LKIF-Core a common sense ontology for law

• Law consists of terms to interpret cases in the
  real (social) world
• Cases are described in common sense terms
• Legal technical terms are usually more strictly
  and explicitly defined common sense concepts
• Eg liability, responsibility, permission, murder,
  etc.
• Legislation and contracts often contain explicit
  definitions of terms ? ontology
• Aligning two vocabularies

29
Dependencies between types of legal knowledge
Legal domain ontology
30
LKIF-Core a common sense ontology for law

• Law consists of terms to interpret cases in the
  real (social) world
• Cases are described in common sense terms
• Legal technical terms are usually more strictly
  and explicitly defined common sense concepts
• Eg liability, responsibility, permission, murder,
  etc.
• Legislation and contracts often contain explicit
  definitions of terms ? ontology
• Aligning two vocabularies

31
there are exceptional domains where common sense
terms cannot be aligned..
32
Top-down development/KA support
LKIF-Upper upper ontology
mental concept
social concept
physical concept
physical object
physical process
mental object
content
intention
role
agent
norm
action
document
organization
anchors
LKIF-Core legal core ontology
legal action
legally valid norm
legal code
legal person
judicial organization
judge
normative article
Dutch penal code (WvSR)
criminal court
perpetrator/ accomplish
crime
article of WvSR
citizen
Is-a
legal domain ontology (Dutch) criminal law
Part-of
33
Common sense

• By definition the knowledge that we all share
  (in a culture)
• Test what can be left unsaid
• tacit knowledge
• Descartes (Discourse de la Methode)
• Nobody complains having a lack of common sense

34
Nobody?
35
An evolutionary perspective on acquiring c.s.
concepts

• Concepts enable an organism to perceive things
  and events as instances of objects, resp.
  processes
• Cultural evolution
• Accumulationstanding on the shoulders of
  giants (Newton)
• Survival passing the empirical reality test
  (Popper, 72)
• Taxonomy of survived knowledge species
• Evolutionary psychology
• It all started in the biology
• instincts (e.g. Pinker, 2008)

36
Distance sensors locomotion

• Space the canvas
• Vertical gravity (kinesthesis)
• Horizontal stereo (f/b l/r)
• Statics
• Objects at positions
• Mass, matter (substance), extension
• Dynamics change
• Events with speeds
• Processes (causes of change)
• Complementary view background/focus
• Hobbs (95, 05)

37
Predators and prey

• Agents
• Causation and intention (threat)
• Cooperation
• Differentiation (sex)
• (proto-) roles

38
Enabling reflection reification by propositions
some higher mammals
semantic representations (preverbal messages)
FORMULATOR
39
The mental world

• Propositional representation
• input mode independent
• interoperability
• Reflection
• Mental life as a metaphor of the physical world
• Mental processes and mental objects

40
The social world

• Configurations of roles
• Dissociating agent from role
• Analogue function of physical object (device)
• Role
• Mental object
• Reciprocal relationships
• transactions
• Prescriptions of behavior
• Prediction by a teleological perspective
• Norms
• Roles are associated with positions

41
LKIF-Core main worlds

• physical world
• mental world
• roles ( social world)
• abstract world
• occurrences (terms to refer to occurrences)

42
physical world

• basic natural concepts energy matter
• basic defined concepts physical object process
• both contain mixtures of energy matter
• processes are changes
• transfer (changing positions)
• changing value (quality quantity)
• transformation (changing type of process or
  object)
• types of processes
• mechanics movement (moving objects oneself)
• thermo-dynamics heat exchange, burning,
  (friction)
• light (radiation)
• chemistry solving/mixing/cooking substances

43
process and object
44
Between death, life and mind

• Biology/life
• Living and moving physical objects agents
• Agents have minds
• Minds contain (memories, ), intentions
• Processes initiated by agents actions
• Awareness communication actions (cf speech acts)
• Self awareness reflection
• Control over reasoning
• Modeling fellow agents
• Modeling discourse
• human minds

45
the mental world as a metaphor of the physical
world

• mappings
• energy --gt emotionmotivation
• matter/substance --gt thought/content
  (information)
• object ---gt mental-object (concept,)
• container ----gt mind, memory
• process ---gt mental-process (thinking,
  memorizing, )
• process --gt action
• mind/body problem
• person has mind mind is container of mental
  entities
• action will as force (energy to load the
  intention)

46
roles

• distinguishing between
• role and role taker e.g. student - person
• roles imply complementary relations
• speaker-hearer, student - teacher
• these complementary relations explain
  duty/rights relations in legal theories
• roles are behavioural pre-scriptions
• requirements for role taking (cf man taking
  mother role)
• norms, prescriptions
• role is subclass of mental object
• role performance may be assessed against role
• Bad cook, good cook,
• violating legal norm
• social organization part-of structure of roles,
  defining social positions

47
abstract concepts

• limited to purely formal, mathematical concepts
• evolutionary starting with count-number (cf
  Lakoff Nunez, 2000), but also point, line, size
  (geometry)
• the concrete vs abstract distinction is
  covered in LRI-Core mainly by physical vs
  mental
• mental objects believe, thought,
• also non formal views on proposition, set, logic,
  rectangle, ..
• mental (epistemological) roles hypothesis,
  evidence,

48
and to be able to talk about occurrences

• entities ((instances of) individual objects)
• events and states of entities (explained by
  processes)
• situations and histories of entities information
  management (episodic memory)
• causation as the glue between events
• on the canvas of space and time
• spatial positions/areas
• temporal moments/durations
• now appears to move by the arrow of time
  existence of objects as trajectories in
  space/time

49
LKIF-Core development

• Common-sense top-ontology (LRI-Core)
• Collecting terms (plm 200) from legal experts (6)
• Ratings (relevance, abstraction level,
  common-sense, etc.)
• Middle out approach identifying clusters ?
  modules

50
LKIF-Core levels and modules
51
You can get LKIF core at

• http//www.estrellaproject.org/lkif-core/
• See also Deliverable 1.4
• Literature
• Rinke Hoekstra, Joost Breuker, Marcello Di Bello,
  and Alexander Boer. LKIF core Principled
  ontology development for the legal domain. In
  Joost Breuker, Pompeu Casanovas, Michel Klein,
  and Enrico Francesconi, editors, Law, Ontologies
  and the Semantic Web, volume 188 of Frontiers of
  Artificial Intelligence and Applications. IOS
  Press, Amsterdam, 2009.