Development and application of an expert system for classification domains : a dissertation submitted in partial fulfillment of the requirements for the degree of Bachelor of Agricultural Science (Honours) in the University of Canterbury [Lincoln College]

Abstract

Expert systems are computer systems that embody organised knowledge about some area of human expertise, which enable them to perform as skilful and cost effective consultants. A history of expert systems development and a discussion of the components has been carried out. This includes inferencing techniques, knowledge representation, and knowledge acquisition.

A discussion of the future that expert systems may have in agriculture is also presented. There are many domains in agriculture where expert knowledge is available yet few systems are in use. There seems to be potential for these systems as tools for agricultural extension and technology transfer.

An expert system shell (ABE) for domains involving classification of individuals or diagnosis of problems has been designed and applied to the domain of grass identification. The basis of this system is a new form of knowledge representation in tables. Table based knowledge representation is conceptually easy to understand and code.

In order to assess the usefulness of the grass identification system as a teaching tool, members of a first year plant science class used ABE to identify a number of grasses. The students reacted to the system very favourably and fifty percent of identifications were correct. It was found that enhancement of the explanation facilities, possibly by including graphics, would increase the ease of use.