Translation from More to Less Expressive Representation Languages

摘要

Knowledge Base reuse is a real problem of great practical significance. Currentlyit is not possible to reuse a knowledge base developed for another purpose. As a result, knowledge-based systems are very expensive to develop because it is always necessary to start from scratch on a new knowledge base. A significant stumbling block to knowledge base reuse is that all practically usable knowledge bases are developed in specialized representation languages. Hence, reuse requires translation between specialized languages. This type of translation is particularly difficult when the target language is less expressive than the source. The goal of this project is to develop a translator shell that enables the rapid development of translators between specialized representation languages. In the first year, a prototype of the shell was developed and feasibility demonstrated with the development of a Knowledge Interchange Format (KIF) to CLASSIC translator. In the second year, an EXPRESS to KIF translator and a LOOM to KIF translator were completed and tested on several example knowledge bases, the translator shell was significantly enhanced to support several new capabilities, experiments were successfully conducted to show that Ontolingua could be significantly improved by replacing its translators with translators developed using our shell, an activity to apply our work to translation between databases was begun. In the third year, a prototype translator between human genome databases was developed, support of the translator shell technology continued, and further research was conducted on the KIF to KIF translation problem.