A Software Environment for Developing Large-Scale Knowledge Bases

Knowledge based methods have the potential to offer high performance solutions to a wide variety of problems. However, the development of large scale knowledge bases (KBs) and their effective application to problems have so far proved to be more expensive and less feasible than other competing methods such as statistical methods or weak search techniques. There is an immediate and critical need to design a software environment for developing large scale KBs that (i) engages developers in knowledge elicitation dialogs, thereby reducing the amount of training required, (ii) automates quality control by implementing KB development guidelines, testing and gap detection routines in automatic and interactive procedures, (iii) aids resource allocation and management of large KB development projects involving many developers; and (iv) speeds up knowledge acquisition through the use of generative acquisition rules. In addition, since large KBs necessarily contain large amounts of information for any query, intelligent routines must be developed to filter out irrelevant information. The proposed Phase I work will address the above needs and carry out the design, feasibility study, and data collection activities necessary to build such a software environment for KB development during Phase II. Phase I work will involve analysis and small scale prototype building.