An inference mechanism under incomplete knowledge based on rule similarity considering viewpoint

This paper describes a mechanism that uses similarities between rules to obtain an approximate solution when the knowledge stored in the problem-solving system is incomplete. When the knowledge base does not contain rules needed to solve a problem, the problem is solved by applying similar rules. The similarities between rules are measuring considering the circumstances of the problem. A large-scale knowledge base of Japanese words, automatically constructed from machine-readable dictionaries, is used to measure the similarity between rules. This mechanism has been implemented by using predicate logic and has been tested experimentally by solving simple problems.