Knowledge Engineering 2018

Knowledge Engineering Environment for Market Analysis and Interpretation
Traditional methods for constructing knowledge-intensive systems have relied heavily on intervention from artificial intelligence specialists. The first step in this intervention is typically either: extensive human-directed interviews of subject-matter experts so that the knowledge can then be laboriously hand-encoded; or the training of SMEs in some highly-restricted intermediate representation. Such systems have been costly to produce, and have typically failed to model expert knowledge to any degree of complexity outside of very narrow domains. This failure has been a major obstacle to the development of systems that harness human reasoning with a computer’s tireless attention to detail. We propose to investigate why knowledge engineers need to be in the loop, and provide requirements and high-level design to addresses many of these problems. Our focus will be on AI interfaces that operate in the SME’s domain of discourse. As a concrete example, we will be looking at a mixed graphical/textual representation of the intelligence analysis process whereby complex rules can be expressed as simple questions in a context inherited from the workflow. Our experimentation to date has found that this representation is not only very intuitive for SMEs, but is also highly productive in comparison to more conventional rule-construction methods.

Potential markets:

Related