@techreport{R-87-18,
TITLE = {Reuse and Maintenance Techniques in Knowledge-based Systems},
AUTHOR = {Henrik Nordin},
YEAR = {1987},
NUMBER = {R-87-18},
INSTITUTION = ida,
ADDRESS = idaaddr,
ABSTRACTURL = {/publications/cgi-bin/tr-fetch.pl?r-87-18+abstr},
ABSTRACT = {As more and more knowledge-based systems are developed, the need for reuse and maintenance of the knowledge bases increases. Compared to traditional software engineering, the situation is improved by the separation of different kinds of knowledge and their explicit representation. Still, problems remain. In particular, it is often unclear how different parts of the knowledge base depend on each other, and on the deeper principles on which they are based. In this report, I claim that this is an important problem in the domains we have studied, and propose a project to study how it can be alleviated. I will use two main approaches. The first tries to find approximate partitions by analyzing both the knowledge in the system and meta-level knowledge, e.g. in the form of justifications. The second approach goes a step further and analyses the derivation of the knowledge in order to know what needs to be changed when a certain piece of knowledge changes. The resulting analysis tools will be integrated into an environment with support for development of technical fault diagnosis systems from reusable causal models to deliverable systems based on faults and symptoms.},
IDANR = {LiTH-IDA-R-87-18}