A common-sense reasoning system for mechanical engineering

Abstract

This thesis describes the Mechnet system, which was created to test an implementation of analogy-based reasoning about mechanical engineering, focusing on tools and machines that would be found in a shop. Mechnet uses common-sense data about how these things are used and what they are made of, and attempts to judge which objects are similar to each other. The goal was to make "sensible" analogies about things, in that they make sense to a human who has experience in this area. Results show some reasonable analogies for many tools and machines entered, but non-sensical analogies in several instances. The problems which cause these non-sensical analogies are discussed and analyzed. Recommendations are made for improvement; and further research into other kinds of common-sense reasoning.