Conceptual Graphs

Contents
• Definition of Conceptual Graphs
• Basic building blocks
• Concept node representation
• Exercise
 Definition of Conceptual Graphs
• John Sowa, formerly of IBM, is one of the key proponents
of conceptual graphs (CG). Sowa’s project is to create "a
system of logic for representing natural language
semantics".

• Conceptual graphs form a knowledge representation

language based on the one hand in linguistics, psychology
and philosophy, and data structures and data processing
techniques on the other.
 Definition of Conceptual Graphs
• The main aim is mapping perception onto an abstract
representation and reasoning system.

• A conceptual graph consists of concept nodes and relation

nodes
– The concept nodes represent entities, attributes, states, and events
–  The relation nodes show how the concepts are interconnected
 Conceptual Graphs: Basic Structure
("The cat sat on the mat")
Rules for assembling
Words
percepts
Percepts Grammar Rules

CAT STAT SIT LOC MAT

PS: percepts are fragments of images that fit together like pieces of a jigsaw puzzle
 Conceptual Graphs: Basic Structure
• Alternative notation for text based representation:

[cat] --> (stat) --> [sit] --> (loc) --> [mat]

• Square brackets denote concept nodes.

• Parentheses denote relation nodes.
 A Graph-Theoretic Definition
• Conceptual Graphs are finite, connected, bipartite
graphs.

– Finite: because any graph (in 'human brain' or 'computer storage')

can only have a finite number of concepts and conceptual
relations.

– Connected: because two parts that are not connected would

simply be called two conceptual graphs.

– Bipartite: because there are two different kinds of nodes:

concepts and conceptual relations, and every arc links a node of
one kind to a node of another kind
 Perception
• ‘Perception is the process of building a working model that
represents and interprets sensory input’.

• The reception of sensory input, ‘a mosaic of percepts’, is

converted into concepts:
– Concrete concepts – that have associated percepts
– Abstract concepts – that do not have any associated percepts.
 Perception
• For Sowa, a sensory icon is matched in an ideal brain to a
single percept or to a collection of percepts, which are
combined to form a complete image: an interconnected set
of percepts.

• Percepts are combined in the brain and their

interconnections stored as a conceptual graph.
 Conceptual Graphs Example
• Consider the sentence: "A cat sitting on a mat"

• This sentence can be interpreted at different levels:

1. There are concrete concepts: cat, mat and sitting which enable us
to experience the external word and motor mechanism to react to
it.

2. The words of our natural language, arranged in accordance with

the grammar of the language, is one way of articulating and
disseminating the experience.
 Conceptual Graphs Example
3. Each of the concepts in the sentence belongs to, or can be related
to, a category or class:

Animal>Cat; Furniture>Mat; Posture>Sit;

Living Being>Animal; Household Objects>Furniture; Act>Posture

Thus

Cat – Sit – Mat Increasing

Animal – Posture – Furniture Abstraction
Living Being – Act – Household Object

A hierarchy of concept type defines the relationship between concepts at

different levels of generality
 Conceptual Graphs Example
4. The concepts cat-sit-mat are related to each other in that:

– It is a common observation that some animate objects do sit on

certain concrete objects

– Even if we had never seen a cat sitting on a mat, we may derive the
conceptual graph on the basis of observation

– The order of the concrete concepts is important in that were we to

say that mat-sit-cat, it would be difficult to match this stated percept
with a conceptual graph in the ideal brain.

– Formation rules determine how each type of concept may be linked

to conceptual relations.
 Conceptual Graphs Example
5. The above sentence relates to an episode or to some context to
which it is relevant.

6. Each episode may have some deeper mental associations, like

emotions.

7. When we ask the question: what is the cat doing?, the answer is
that the cat is sitting and that its current location is the mat. The
cat’s STATe, its current ACTivity, its LOCation may each be
related to a procedure of some type.
 Conceptual Relations
• Concepts are linked by conceptual relations to form a
conceptual graph.

• If a conceptual relation has n-arcs, then it is said to be n-

adic, and its arcs are labelled 1, 2, …..n
 Example
• Consider the sentence:
– Mary gave John the boring book authored by Tom & Jerry

(1) (2) (3)

• There are three main parts: (1), (2), and (3)

 Example
(1): Mary gave John the boring book authored by Tom & Jerry

Person: Mary agent give

Person: John recipient

Both relation nodes have two arcs each and are referred to as expressing a 2-
ary or binary relation between the two concepts
 Example
(2): Mary gave John the boring book authored by Tom & Jerry

book boring

The relation node has only one arc and thus refers to a 1-ary or unary relation
 Example
(3): Mary gave John the boring book authored by Tom & Jerry

Person: Tom

book author

Person: Jerry

The relation node has 3-arcs and is referred to as expressing 3-ary or ternary
relation
 Formal Conceptual Relations

Concept 1 Concept 2 Relation

Entity:*x Entity*y accompaniment (ACCM)
    attribute (ATTR)
    characteristic (CHRC)
    content (CONT)
    part (PART)
    possession (POSS)
support (SUPP)
Event(Act) Attribute manner (MANR)
 Formal Conceptual Relations

Concept 1 Concept 2 Relation

Event(Act) Entity result (RSLT)
    source (SOUR)
Event(Act) Entity (Animate) agent (AGNT)
    recipient (RCPT)
Event(Act) Entity (Place) destination (DEST)
    path (PATH)
  Entity (Substance) material (MATR)
Function Data argument (ARG)
State*x State*y causation (CAUS)
 Concept Nodes
• Recall that in the discussion of Collins and Quillian’s
semantic networks, we have found that these networks
were logically inadequate!

• This situation was not resolved in some of the subsequent

formulations of semantic networks. Specifically, it was
difficult in a typical semantic network notation to
distinguish between nodes describing:
– classes and subclasses
– classes and members
 Concept Nodes
• In the sentence:
– Tom is a cat, a feline mammal

Tom is_a cat is_a feline is_a mammal

individual species subclass class

• The relation "is_a" is used to describe relationships

between concepts that are mildly different.
 Concept Nodes
• A good representation should allow us to distinguish
between:
– Individuals and species
– Species and classes
– Classes and subclasses

• Individuals may have properties that may not influence

their belonging to a subclass:
– Tom is a brown tabby

• Should not influence the observation that:

– A tabby cat is a kind of cat
 Concept Nodes
• In CG theory, 'every concept is a unique individual of a
particular type'.

• Concept nodes are labelled with descriptors or names like

"dog", "cat", "gravity", etc. The labels refer to the class or
type of individual represented by the node.

• Each concept node is used to refer to an individual

concept or a generic concept.

• In CG theory we have a relation called: name

 Concept Nodes
• CG allows nodes to be labelled simultaneously with the
name of the individual the node represents and its type.
The two are separated by a colon (":")

• Consider the example:

– Tom, a cat, is brown

cat: "Tom" colour brown

 Concept Nodes: Unnamed Individuals
• Consider the example that we do not know the name of a
cat that is brown:

cat: #12345 colour brown

• Each concept node in a CG may be used to represent

specific but unnamed individuals by a unique prescribed
number.
 Concept Nodes: Multiple Names
• We subsequently found out that the cat is called by
different names: "Sylvester", "Sugar Pie" and "Squidgy
Bod":

name "Sylvester"

cat: #12345 name "Sugar Pie"

name "Squidgy Bod"

 Concept Nodes: Unspecified Individuals
• General markers can also be used to refer to an unspecified
individual. The CG:

cat colour brown

• Refers to an unspecified cat. Notationally, unspecified

individuals are shown by the existence of an asterisk ("*")

cat: * colour brown

• BUT… this is usually omitted (cat = cat:*).

 Concept Nodes: Named Variables
• Named variables can also be used to refer to an individual. These are
represented by an asterisk followed by the variable name.

• This is useful to indicate nodes that are the same unspecified

individual.

dog:*X agent scratch object ear

instrument part

paw part dog:*X

 Canonical Graphs
• A conceptual graph is a combination of concept nodes and relation
nodes where every arc of every conceptual relation is linked to a
concept. This could lead sometimes to sensible statements like
– "a bunny sitting on a mat"
and at time will lead to nonsense like:
– "colourless green ideas sleep furiously"

• Sowa distinguishes the nonsensical graphs from those that represent

real or possible situations in the external world by declaring the later
as canonical.

• Certain conceptual graphs are canonical. New graphs may become

canonical or be canonised by perception, formation rules, or through
"insight".
 Exercises
• Please create the conceptual graph of the following
sentence:
– John is between a rock and a hard place
 Solution 1
• "John is between a rock and a hard place"

rock

person: John between

place

attribute hard