06.03.2013 Views

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

Artificial Intelligence and Soft Computing: Behavioral ... - Arteimi.info

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

autonomous learning <strong>and</strong> refining knowledge from the external world. One<br />

main difficulty with manual acquisition of knowledge is that the experts often<br />

fail to correctly encode the knowledge, though they can easily solve a complex<br />

problem of their domain. Further, the ordering of the pieces of knowledge<br />

carried out by the experts, being sometimes improper, causes a significant<br />

degradation in search efficiency of the inference procedure. Lastly, the<br />

certainty factor of the pieces of knowledge, set by the experts, too, is not free<br />

from human bias <strong>and</strong> thus may lead to inaccurate inferences. The need for<br />

automated knowledge acquisition is, therefore, strongly felt by the scientific<br />

community of AI.<br />

20.2 Manual Approach for<br />

Knowledge Acquisition<br />

Knowledge acquisition is a pertinent issue in the process of development of<br />

expert systems. A good expert system should contain a well-organized,<br />

complete <strong>and</strong> consistent knowledge base. An incomplete or inconsistent<br />

knowledge base may cause instability in reasoning, while a less organized<br />

system requires quite a significant time for search <strong>and</strong> matching of data. The<br />

malfunctioning of the above forms originates in an expert system generally<br />

due to the imperfections in i) the input resources of knowledge <strong>and</strong> ii) their<br />

encoding in programs. The imperfection in the input resources of knowledge<br />

can be overcome by consulting proved knowledge-rich sources, such as<br />

textbooks <strong>and</strong> experts of respective domains. The encoding of knowledge<br />

could be erroneous due to either incorrect underst<strong>and</strong>ing of the pieces of<br />

knowledge or their semantic misinterpretation in programs. A knowledge<br />

engineer, generally, is responsible for acquiring knowledge <strong>and</strong> its encoding.<br />

Underst<strong>and</strong>ing knowledge from experts or textbooks, therefore, is part of his<br />

duties. A clear underst<strong>and</strong>ing of the knowledge base, however, requires<br />

identification of specific knowledge from a long narration of the experts. The<br />

knowledge engineer, who generally puts objective questions to the expert,<br />

therefore, should allow the expert to answer them in sufficient detail,<br />

explaining the points [6]. The semantic knowledge earned from the experts<br />

could be noted point-wise for subsequent encoding in programs. Occasionally,<br />

the experts too are not free from bias. One way to make the knowledge base<br />

bias-free is to consult a number of experts of the same problem domain <strong>and</strong><br />

take the view of the majority of the members as the acquired knowledge.<br />

20.3 Knowledge Fusion from<br />

Multiple Experts<br />

Fusion of knowledge from multiple number of experts can be implemented<br />

in either of the following two ways. First the knowledge engineer may

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!