The Inference of Domain Structure from Informal Process Descriptions
Abstract
Understanding informal descriptions of processes requires access to a body of knowledge about the process domain, and the ability to use that knowledge appropriately. A great deal of effort has been spent in developing methods for organizing and using domain knowledge; relatively little has been done to automate acquisition of such knowledge. Since English process descriptions reflect the underlying structure of the process domain, knowledge about that structure may be inferred from the description itself. A categorization of important structural knowledge classes is presented, and a production system described which interprets English-like statements on the basis of existing structural context. A sample of the rules from this system is examined. By assuming conditions required in the rule patterns when a linguistic structure is not interpretable, it is possible to infer a great deal of structural knowledge about a process domain. This incremental growth of domain structure presents an alternative to constructing process understanding systems applicable only to very restricted domains, or requiring extensive additions of domain-specific knowledge by human experts for each new task.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1977
- Accession Number
- ADA048154
Entities
People
- David Wile
- Neil Goldman
- Robert Balzer
Organizations
- University of Southern California