Acquiring and Filtering Knowledge: Discovery & Case Based Reasoning.

The acquisition of knowledge for the development of knowledge-based reasoning systems has been source of interest in artificial intelligence for some time now. The use of experts as the main source of knowledge in the development process has put some strain on the expert system development cycle because of the difficulties associated in the articulation of the experts' largely tacit knowledge into discernible knowledge structures like rules and frames. A specialist domain, characterised by a largely restricted vocabulary and typified by the presence of conventions required for consensus amongst members along with the abundance of text repositories, presents us with an opportunity to determine if a method can be developed which is capable of automatically extracting knowledge from text documents. It is our hope that the presence of this restricted language will result in the presence of recurring linguistic patterns that can be extracted and hence determined as templates of use. We have in this investigation, analysed journal articles in the area of carbon nanomaterials with emphasis on experimental papers which show similarities to recipes and hence can easily be converted into cases for use in a case-base reasoning system. We have used techniques of corpus linguistics to analyse a collection of text and determine the existence or not of these templates. We have determined some fairly frequent patterns of use which have been used in the development of a template for the extraction of objects and their corresponding values for use in a case-base. These patterns of use are an indication of the ontological commitment of the writers within the domain