Knowledge-based understanding on a small machine

One problem plaguing current knowledge-based systems is the acquisition of the information that is represented in the knowledge base. Most knowledge processing systems, such as expert systems and more recent case-based systems, are either based on static data that has already been entered into the system, or use human intervention to enter the information into the system in the correct format. Both of these methods greatly reduce the power and flexibility of the system since, presumably, an intelligent knowledge-based system would make better decisions provided with more information. One resource that provides a vast amount of information is written text, such as journals, newspapers and on-line news wires. Systems that use information provided by such resources could greatly benefit if they automatically acquired this information. This paper presents research directed at the development of such a system on a small machine.