Semantic_net_and_Frames_in_knowledgeR.pdf

Topics covered:
knowledge representation
Semantic networks
frames

1. Knowledge Representation
By:
Vijeta Rani
Delhi Technological University
© Delhi Technological University

2. Meaning of Knowledge Representation
"Knowledge is defined as the Intellectual acquaintance with, or
perception of, facts or truth.
"A representation is a way of describing certain fragments of
information, so that, any reasoning system can easily adopt it for
inferencing purposes.
"Knowledge representation is a study of ways of how knowledge is
actually picturized and how effectively it resembles the representation
of knowlege in human brain.

3. "One can represent information about an object or event by means of a
database and manipulated by a DBMS.
"A DBMS, even though, holds information, do not hold the facility for
representation and manipulation of facts like:-
1. All carnivores have sharp teeth,
2. Cheetah is a carnivore.
3. Hence Cheetah has sharp teeth.
"The third statement can be inferred by a human mind or a AI s Knowlege
base management system. But in DBMS, this information is needed to
be explicitly specified.
Limitations of DBMS

4. Database
"Collection of data represting facts.
"Large volume of data and facts
"Change over time
"Operates on a single object.
"Updates are performed by clerical
personnel.
"All information needed to be
explicitly stated.
"Maintained for operational
purposes.
"Represented by relational or
network or hierarchical model.
Knowledge Base
"Has information at a higher level of
abstraction
"Significantly smaller than DB
"Changes are gradual.
"Operates on a class of objects
"Updates are performed by domain
experts
"Has the power of inferencing.
"Used for data-analysis and planning.
"Represented by logic or rules or
frames or semantic nets.

5. Mapping between facts and representation
Facts are truth in real world which we want
to represent They are represented in some
choose formalism which can be manipulated
by computer which is called knowledge. They
are represented in terms of symbols
Lucy is a dog
Dog(Lucy)
All dogs have tail
for all x, Dog(x)!’hastail(x)
Hastail (Lucy) !’Lucy has tail
A good system for the Representation of
knowledge in a particular domain should
posses the following property: Well defined
syntax, semantic and capability to express
complex sentences

6. Properties of a good knowledge representation System
Representational Adequacy : The ablity to represent all kind of knowledge

Inferential Adequacy : The ability to manipulated the represent
knowledge to derive new knowledge from old knowledge.

Inferential Efficiency : Should be able to in corporate additional
information that can be used to inference new knowledge from old
knowledge

Acquistional Efficiency : The ability to acquire new information easily,
direct insertion by person in to data base or through program .

7. Approaches to Knowledge Representation
1. Simple Relation Knowledge : The simplest way to represent
declarative facts is as a set of Relation of the same form as used
in database .

8. 2. Inheritable knowledge : Objects must be organized into classes and
classes must be arranged in a generalized hierarchy. There is the property
of inheritance.

9. 3.Inferential knowledge: Traditional logic is used to represent knowledge
and inference procedure such as resolutions which would infer from given
knowledge. Reasoning is done forward or backward.
4. Procedure knowledge: Operational or procedural Knowledge that
specifies what to do when.

10. Declarative Representation
"It declares every piece of knowledge
and permits the reasoning system to
use the rules of inference to come out
with new pieces of information.
"E.g.
Vx (Carnivore(x) !’ Sharp teeth(x))
Carnivore(Cheetah)
Cheetah !’ Sharp teeth
Procedural Representation
"It represents knowledge as procedures
and the inferencing mechanisms
manipulate these procedures to arrive
at the result.
"E.g.
procedure carnivore(x)
If (x==Cheetah) return TRUE;
Else return FALSE;
end procedure carnivore(x)
procedure sharp_teeth(x)
If carnivore(x) return TRUE;
Else return FALSE;
end procedure sharp_teeth(x)

11. Declarative Representation
"This approach is flexible.
"Modularity is higher in this, because
each piece of knowledge is an
independent chunk on its own.
"It is enough to represent the
knowledge only once, because it can
store knowledge in the form of
variables.
Procedural Representation
"In this, heuristic knowledge can be
easily represented.
"One has the control over search, which
is not available in declarative
knowledge representation.
"In this approach also, variables can be
used to store knowledge.

12. Issues in Knowledge Representation
* Are there any attribute of objects so basic that they should be covered?
* Are there any important relationship among attribute of objects?
* At what level knowledge be represented? Is there any good set of
primitives?
* How should set of objects be represented?
* How can relevant part be accessed when large knowledge is stored in
knowledge base?

13. Some Knowledge Representation Schemes
"Semantic nets
"Frames
"Conceptual dependency
"Scripts

14. Semantic Networks (Net)

15. History of Semantic net
" Semantic Nets were first invented for computers by Richard H. Richen
of the Cambridge Language Research Unit in 1956 as an interlingua
for machine translation of natural language. They were developed by
Robert. F. Simmons as System Development Corporation, California in
the early 1960s. Later improved in the work of M.Ross Quilian in 1966 to
use as a modal of human mind.
"Computer implementations of semantic networks were first developed
for artificial intelligence and machine translation, but before using in
computers and machine language it have long been used in philosophy
and psychology.

16. Semantic Networks (Net)
"Semantic net is one of the graphical representation of knowledge.
"In the semantic net, the information is represented as a set of nodes
connected to each other by a set of labeled arcs, which represent the
relationship among them.
"Nodes represent the objects and descriptive information about those
objects. They can be Entities, Attributes, states or events.
"Links describe the relationships between the nodes and labels on it
specify the type of relationship.

17. Semantic Networks (Net)
"Object can be any physical item such as book, car, desk, or even a
person.
"Objects can also be concepts, events, or actions.
"Attributes of an object can also be used as nodes. These may represent
size, color, class, age etc.

18. Example of a simple semantic net

19. From the previous Semantic Net, it is possible for us to say that :
1.A scooter is a two wheeler and it is a moving vehicle.
2.A moving vehicle needs an engine (could be petrol or diesel or any
engine), a fuel engine system to sustain the engine running, an electrical
system for its lights , horns and also for initial ignition ( in case of petrol
vehicles ) and brakes.
Such semantic net not only gives details about an object under
consideration but also provides facilities to represent variables. This
semantic net has two variables X and Y as a part of the node. This
implies that the speed of the printer could be 300 or 600 lines per minute
and DTU could have 18 or 25 or 30 departments. Figure on next slide.

20. Representation of variables in sementic network

21. Classification of Nodes in a Semantic Net
The nodes in the semantic net are classified as:
1. Generic Nodes: It is a very general node. In the semantic network on
previous slide, for the semantic network of DTU centre, the printer is a
generic node because many printers exist and that node has to cater to all
of them.
2. Individual or Instance nodes : Individual or instance nodes explicitly
state that they are specific instances of a generic node. HP deskjet Printer
could be an individual node because it is a very specific instance of the
Printer.

23. Links in a Semantic Net
Many link structures are being used in semantic nets , some of them are:
1. is_a
2. has_a
3. has_dept
4. contains
5. part_of
6. is_in
etc.
Of all these links, is_a link is a special type of link because it provides
facilities to link a generic/individual node to a generic node.

24. is_a Link
ØMajor feature of a is_a link is that it generates hierarchal structure
within the network.
ØIs_a link has another major property which is called as inheritance . The
property of inheritance is that the properties which a most a generic
node possesses are transmitted to various specific instances of a
generic node. This property is called transitive property of inheritance.
ØAn individual or instance node forms a subset of another generic node
etc.

26. "In the previous slide, the properties of the vehicle are also applicable to
the leaf nodes. For example every vehicle has to have an engine for
moving which holds good for either Kawasaki Bajaj or Challenger .
Also that a vehicle s purpose is for transportation which is true either for
an Eicher Mitsubishi or TGV of France. This property of inheritance
helps in jumping from one level to another carrying the characteristics
of the generic node to very specific instances.

27. Basic types of links-relationship
"Is-a link: It represents the inclusion relationship of an object in
another(i.e. to link a class and its superclass).Ex: Bird isa mammal.
" Has-part link: An object is described by another object(Bird has-part
Wings and Bird has-part Feather)
"Instance link: It represents the relationship between a type and a token
(sparrow is an instance of bird)
"A node can have any number of superclasses that contain it.
"A node can be inherited by the properties of multiple parent nodes and
there ancestors in the network. It can cause the conflict inheritance.

28. Example

29. Example

30. Example

31. Reasoning using semantic networks
Reasoning steps in semantic nets:-
1.Specify the Start Node.
2.From the initial node, other nodes are pursued using the links until the
final node is reached.
E.g.
To illustrate this, consider semantic network of DTU Computer Centre, If
one wishes to find What is the speed of the printer?
1. identify the arc that has the characteristics Speed and
2. Find to what node does arc points to.

32. In our example we had given it a variable value (X) which could be a
numeric one. This type of arriving at results by matching nodes and arcs
and the utilization of inheritance property helps to a great extent in the
reasoning process.
The major hurdle in utilizing semantic networks is that there is no
standardization and formalization as far as notations and reasoning is
concerned.
But the overall concept of arcs and nodes in semantic networks has been
standardize

33. Non-binary predicate representation
"Binay predicate logics: Semantic nets are the natural way to represent
the relationship of binary predicates in predicate logic.
"Some Binary predicate logics from last example are:
Isa(Person,Mammal)
Instance(Sachin,Person)
Uniform color(Sachin,Blue)
"Unary predicate logics:They can also be represented in the binary
predicates by using general purpose predicates, like isa and instance.
"Example:
mammal (person) can be represented as isa(Person,Mammal)
"Now its easy to represent it in the semantic net.

34. Non-binary predicate representation
"Three or more predicate logics: Three or more place predicates can also
be converted to the binary form by creating one new object
representing the entire predicate statement.
"Example: score (INDIA ,USA ,7-2)
"It can be represented in the semantic net by creating an another node
represent the specific game and then by relating the pieces of
information as shown in the next slide.

36. Example: John gave the book to Mary.

37. Example

38. In this network, inheritance has been used to derive additional relation.
has_part(Tendulkar,Brain)
INTERSECTION SEARCH: Semantic nets can be used to find relationships
among objects by spreading activation to and from each of two nodes.
Q.What is the relationship between Tendulkar and Blue?
Represention: Semantic nets are the natural way to represent relationships that would
appear as ground instances of binary predicates in predicate logic.
is_a( person, Mammal)
instance(Tendulkar,Person)
Predicates can be thought of as a binary predicates using some general purpose predicates:
person (Tendulkar) can be instance(Tendulkar,person).
Then three or more predicates can be converted to a binary form by creating one new
object representing the entire predicate statement.

39. Conceptual Graphs
"Conceptual graphs are semantic nets representing the meaning of
(simple) sentences in natural language
"It is a technique for representing the content of a declarative sentence
describe the several aspects of a particular event.
"It contains the two types of nodes
Concept node
Relation node( binary relation between concepts)

40. Example

41. Advantages of Semantic Networks
"It is very easy to visualize.
"Abstract knowledge can be represent by linking them together.
"Attributes can be inherited for any object very easily.
"Unary , binary, and more place predicate can be easily represented
through semantic net.
"It is efficient in the space requirements
Object represent only once
"Relationship handled by the arc (pointers).

42. Disadvantages of Semantic Networks
"Facts placed inappropriately may cause the misconception.
"There is no standard about node and arc values.
"Multiple inheritance (Particularly from multiple sources when inheritance are
wanted) can cause the conflict.
"There is no standard definition of link and node names. This make it difficult to
understand the network that s why it is not designed in the consistent manner.
"Initially semantic network was proposed as a modal of human associative memory.
But the human brain contain 10
10
neurons and 10
15
links. Consider how long it take
for the human to answer NO to a question Are there tree on the moon?
Obviously human beings process information in a very different way. It is not as
modeled by the semantic net.

43. Frames
"Frames are a means of representing common sense knowledge.
"Knowledge is organised into small packets called Frames.
"The contents of the frame are certain slots which have values.
"All frames of a given situation constitute the system.
"Whenever, one encounters a situation, a series of related frames are
activated and reasoning is done.
"A frame can be defined as a data structure that has slots for various
objects and a collection of frames consists of expectations for a given
situation.

45. Declarative and Procedural frames
"A frame structure provides facilities for describing objects, facts about
situations, procedures on what to do when a situation is encountered.
"Frames are used two represent two types of knowledge.
1. Declarative/Factual
2. Procedural
"Declarative frames: A frame that merely contains description about
objects.
"Procedural frames: To represent procedural knowledge in a frame by
attaching slots, which explain how to perform things. Such frames are
also called as action-procedure frames.

46. Slots in an Action frame
"Actor slots: They hold information about who is performing the action.
"Object slots: This slot has information about the item to be operated
on.
"Source slot: This holds information that from where the action has to
start.
"Destination slot: This holds informationabout the place where the
action has to end.
"Task slot: This generates the necessary sub-frames required to perform
the operation.

48. Linking of Sub-Frames
"It is possible to extend each frame by associating some other frames to
these. Thus the property of inheritance can be implemented using
frames also, in which, root frame is the generic frame while leaf frames
are specific instances of the root frame.
"Attaching a frame to another, results in a network of frames.
"It is possible to link subframes of both declarative frames as well as
procedural/action frames.

51. Reasoning using Frames
"Reasoning using frames is done by instantiation.
"Instantiation process begins when the given situation is matched with
frames that are already in existence.
"The reaoning process tries to match the frame with the situation and
latter fills up slots for which values must be assigned.
"The values assigned to the slots depict a particular situation and by this,
the reasoning process tries to move towards a goal.
"The reasoning process can be defined as filling frame slots in frames.
"Generally, if a given slot characteristic is not present, the slot provides a
default value for that characteristic.

52. Reasoning using Frames Contd..
"It has been observed that situations are not generally static. If there is
any deviation in the characteristics, the values of the corresponding
slots are updated so that a clear picture is arrived at, that conforms to
that current situation.
"Reasoning using frames allows one to move from one frame to another
to match the current situation. This process builds up a wide network of
frames, thereby facilitating one to build a knowledge base for
representing knowledge about common sense.
"The main task of action frames is to provide a facility for procedural
attachment and help transforming from initial to goal state. It also helps
in breaking the entire problem into sub tasks which can be desribed as
top down methodology.

53. References
"Artificial Intelligence by Elain Rich, Kevin Knight and Shivashankar B Nair.
"Foundations of Artificial Intelligence and Expert Systems by V S
Jankiraman, K Sarukesi & P Gopalakrishan.