Mixed-Initiative Development of Knowledge Bases
Gheorghe D. Tecuci, Mihai Boicu, KATHRYN J. WRIGHT, Seok Won Lee · 1999
Ample experimental evidence shows that manual solutions to the problem of building knowledge bases are highly inefficient, while fully automated, machine learning based, approaches have not yet led to practical solutions. This makes this problem an ideal test bed for developing mixedinitiative methods. In this paper we present such a mixedinitiative approach where a subject matter expert teaches his or her expertise to a learning agent. We present several mixed-initiative methods used during the various phases of knowledge base development, such as, cooperative problem solving, rule learning, rule refinement and exception handling, discussing solutions to the control and communication issues that allow the realization of several types of synergism between the subject matter expert and the learning agent. We also provide experimental evidence of the feasibility of the proposed approach.