Knowledge Acquisition, Induction

Knowledge Acquisition

The success of the knowledge-based systems lies in the quality and extent of the knowledge available to the system. The acquiring and validating a large group of consistent and correlated knowledge is not a trivial problem. This has given the acquisition process a very important role in the design and implementation of these systems. Consequently, the effective acquisition methods have become one of the principal challenges for the AI researchers.

The knowledge acquisition is the process of adding a new knowledge to the knowledge base and refining it or otherwise improving the knowledge that was previously acquired. The acquisition is usually associated with some purposes such as expanding the capabilities of a system or improving its performance at some specified task. It is a goal-oriented creation and the refinement of knowledge. It may consist facts, rules, concepts, procedures, heuristics, formulas, relationships, statistics or other useful information. The sources of this knowledge may include one or more of the following.

• Experts in the domain of interest
• Text Books
• Technical papers
• Databases
• Reports
• The environment.

In order to be effective, the newly acquired knowledge should be integrated with the existing knowledge in some meaningful way so that the non-trivial inferences can be drawn from the resultant body of knowledge. The knowledge should be accurate, nonredundant, consistent that is noncontradictory and fairly complete in the sense that it is possible to reason about many of the important conclusions for which the system was intended in a reliable way.

Induction

The fundamental basis for the methods of decision tree induction or rule induction is that two cases with the similar features which are classified into the same class. The success of such an approach relies on the sufficient training cases in order to cover every aspect of the problem domain without the inconsistency. Otherwise, the induced knowledge may be unreliable. However, in the practical applications, this assumption is commonly violated. In general, the sources of training cases are obtained from a domain expert or from historical data in a well-maintained database. The various difficulties arise typically because the domain expert can provide only a few selected examples whereas the historical data may be obsolete or contain errors and missing values if they exist.

The further difficulty with inductive learning is in the explanation capability. Explaining the reasoning process by which a conclusion is reached is a key requirement for an expert system. However, the rules induced from the training set may differ from those used by the expert. This makes the explanation sometimes less acceptable to the human beings because the underlying process of reasoning that is used by the expert system may be incomprehensible to the users. Thus it is useful to have interactions between the human experts and the inductive learning algorithms. This allows the human experts to provide useful knowledge such as hand-crafted tutorial examples, rules of thumb, general hints and problem-solving strategies in order to help an inductive algorithm.

The inductive algorithm presented is an example that supports the interaction with experts during its learning process. After a knowledge structure is induced from its training cases, the algorithm identifies those cases that cannot be correctly classified. These cases are brought to the expert in the form of questions that are to be solved. Once they are solved they become new training cases to the inductive algorithm for further learning. In this way, the expert knowledge is incrementally elicited and incorporated into the induced knowledge structure.

References:

1. Elaine Rich, Kevin Knight 1991, "Artificial Intelligence".
2. Nilsson, Nils J. Principles of Artificial Intelligence, Narosa Publishing House New Delhi, 1998.
3. Norvig, Peter & Russel, Stuart Artificial Intelligence: A modern Approach, Prentice Hall, NJ, 1995
4. Patterson, Dan W. Introduction to Artificial Intelligence and Expert Systems, Prentice Hall of India Private Limited New Delhi, 1998.