In this article we will discuss about:- 1. Introduction to Knowledge Representation Scheme 2. Metrics of Knowledge Representation Scheme 3. Categories.
Introduction to Knowledge Representation Scheme:
Knowledge Representation (KR) originated as a sub-field of Artificial Intelligence (AI). In the early days of AI, it was sometimes imagined that to endow a computer with intelligence it would be sufficient to give it a capacity for pure reasoning; it quickly became apparent, however, that the exercise of intelligence inevitably involves interaction with an external world, and such interaction cannot take place without some kind of knowledge of that world.
Thus, it became clear that part of the quest for AI must involve the development of methods for endowing computer systems with knowledge. This in turn brought to the fore the question of how such knowledge is to be represented within the computer.
This question can be approached in many different ways, but one can broadly distinguish between approaches which seek to discover, and thereby emulate, the forms in which knowledge is represented in the human brain, and those which take their inspiration from the external forms of representation used by humans to encode their knowledge notably language, mathematics and formal logic.
ADVERTISEMENTS:
The term Knowledge Representation (KR), when used in the AI context, is generally taken to refer to approaches of the latter kind rather than the former, which are regarded as more within the province of Cognitive Science.
We thus find that KR is characterised in the literature in terms which emphasise the quest for explicit symbolic representations of knowledge which are suitable for use by computer some views are:
Knowledge representation and reasoning is the area of Artificial Intelligence (AI) concerned with how knowledge can be represented symbolically and manipulated in an automated way by reasoning programs.
Knowledge representation is the study of how to put knowledge into a form that a computer can reason with…
ADVERTISEMENTS:
Knowledge representation research studies the problem of finding a language in which to encode that knowledge so that the machine can use it.
Knowledge Representation is no longer, however, exclusively the preserve of AI, as witnessed by the following remark by Sowa:
Today, advanced systems everywhere are performing tasks that used to require human intelligence… As a result, the AI design techniques have converged with techniques from other fields, especially database and object-oriented systems.
In particular, Knowledge Representation is closely allied with Formal Ontology, which is concerned with the systematic enumeration and classification of the various kinds of entity represented within a given conceptualisation of the world, together with an account of their properties and relationships.
ADVERTISEMENTS:
Having started life as a discipline within Philosophy, Formal Ontology has become an important strand within information systems research, with particular application to the problems of maintaining coherence and consistency when combining large bodies of knowledge from different sources, as happens ever more frequently with the expansion of the World Wide Web.
Metrics of Knowledge Representation Scheme:
1. Ease of Representation:
The ease with which a problem can be solved depends upon knowledge representation scheme. For example, imagine we have a 3 x 3 chess board with a knight in each corner and we want to know (Fig. 6.1) how many moves will it take to move knight round the next corner.
Looking at diagrammatic representation of Fig. 6.1., the solution is not obvious, but if we label each square and present valid moves as adjacent points on a circle (Fig. 6.2) the solution becomes more obvious; each knight takes two moves to reach its new position so the minimum number of moves is eight.
2. Granuality of Representation:
Granuality of knowledge representation can effect its usefulness, that is, how detailed the knowledge need to be represented. This will depend on the application and the function to which the knowledge will be put. For example, if a knowledge base about family relationships is to be built and we start with ‘cousin’.
We may represent the definition of the relation ‘cousin’ as follows:
ADVERTISEMENTS:
Your cousin is a child of sibling of your parent. When we want to know the gender of the cousin this information may not be enough. For a female cousin your cousin is a daughter of a sibling of your parent and for a male cousin your cousin is a son of sibling of your parent. If we want to know to which side of the parents your cousin belongs we need more information than that our cousin is a child of a sibling of our parents.
A full description of all the possible variations is given below:
i. Your cousin is a daughter of a sister of your mother,
ii. Your cousin is a daughter of a sister of your father,
iii. Your cousin is a daughter of a brother of your mother,
iv. Your cousin is a daughter of a sister of your father,
v. Your cousin is a son of a sister of your mother,
vi. You cousin is a son of a sister of your father,
vii. You cousin is a son of a brother of your mother, and
viii. You cousin is a son of a brother of your father.
It concerns the level of detail with which information is recorded. Most obviously, this could be a matter of resolution; e.g., a map which records the presence or absence of a particular species of plant in each 2 km square has a finer granularity than one which is based on 10 km squares. Similarly a road map which only shows the main trunk roads has a coarser granularity than one which shows all the minor roads in addition. And a map of India showing the state boundaries only has a coarser granularity than that of states.
This is not just a matter of scale; in general the larger the scale of a map, the finer its granularity is likely to be but this correlation is not exact. An approach to a formal treatment of granularity in Al is proposed by Hobbs (1985). Granularity affects both time and space and the two in combination. Granularity related problems arise in a specific form in cartography in relation to map generalization, that is, the task to reducing the level of detail in a map while preserving the correct relationship amongst the features which remain.
3. Property Inheritances:
World can be represented by facts and facts are related to each other. A fact may be specific instance of another more general fact. “Spotty dog barks” for example is a specific instance of the fact “all dogs bark”. This is a case of property inheritance, in which properties of attributes of the main class are inherited by instance of that class. So we represent the knowledge that “dogs bark” and that “spotty is a dog”, allowing us to deduce by inheritance, the fact that “spotty dog barks”.
4. Expressiveness:
An expressive representation scheme will be able to handle different types and levels of granularity of knowledge and knowledge structures and the relationships between them. It will have means of representing specific facts and generic information (say by using variables). Expressiveness also relates to the clarity of the representation scheme. Ideally the scheme should use a notation which is natural and usable both by the knowledge engineer and the domain expert.
Schemes which are too complex for a domain expert to understand can result in incorrect knowledge being held, since the expert may not be able to critique knowledge adequately, that is knowledge representation should be characterised by completeness and clarity.
5. Effectiveness:
In order to be effective the scheme must provide a means of inferring new knowledge from old. It should also be amenable to computation, allowing proper use of adequate tool support.
6. Explicitness:
A good knowledge representation scheme should be able to provide an explanation of its inference and allow justifications of its reasoning. The chain of reasoning should be explicit.
7. Efficiency:
The scheme should not only support inference of new knowledge from old but must do so efficiently in order for new knowledge to be used. In addition, scheme should facilitate efficient knowledge gathering and representation.
Many highly expressive representations are too inefficient for use in certain classes of problems. Sometimes, expressiveness must be sacrificed to improve efficiency. This must be done without limiting the representation’s ability to capture essential problem-solving knowledge. Optimising the trade-off between efficiency and expressiveness is a major task for designers of intelligent systems.
Categories of Knowledge Representation Schemes:
Over the past 40 years, numerous representational scheme have been proposed and implemented, each of them having its own strength and weakness.
According to Mylopoulos and Levesque (1984) they have been classified into four categories:
1. Logical Representation Scheme:
This class of representation uses expressions in formal logic to represent a knowledge base. Inference rules and proof procedures apply this knowledge to problem solving. First order predicate calculus is the most widely used logical representation scheme, and PROLOG is an ideal programming language for implementing logical representation schemes.
2. Procedural Representation Scheme:
Procedural scheme represents knowledge as a set of instructions for solving a problem. In a rule-based system, for example, an if then rule may be interpreted as a procedure for searching a goal in a problem domain: to arrive at the conclusion, solve the premises in order. Production systems are examples of a procedural representation scheme.
3. Network Representation Scheme:
Network representation captures knowledge as a graph in which the nodes represent objects or concepts in the problem domain and the arcs represent relations or associations between them. Examples of network representations include semantic network, conceptual dependencies and conceptual graphs.
4. Structured Representation Scheme:
Structured representation languages extend networks by allowing each node to be a complex data structure consisting of named slots with attached values. These values may be simple numeric or complex data, such as pointers to other frames, or even procedures.