Machine-negotiated, ontology-based EDI (Electronic Data Interchange)

Abstract

• 1. EDI is held back by inflexibility and complexity for users. This is partly due to the antiquated X12 and EDIFACT standards.

• 2. Truly automatic translation between two disparate databases, or between EDI and a database not set up just for EDI, requires machine representation of the concepts and meaning of the data schema, not just the schema.

• 3. The database world has reluctantly begun to recognize this, and that common ontologies are needed for true automated integration. The EDI community, due to its particular culture, may never realize it.

• 4. The real-world “common ontology” is actually the most valuable thing about EDIFACT and X12, but it is informal, in English, and unavailable for any computational use.

• 5. The current major Artificial Intelligence efforts to build large generic ontologies should be applied to automate EDI translations and do useful inference; also, AI ontologists can exploit thousands of practical real-world concept-categories from EDI standards EDIFACT and X12. These provide good target concepts for us to define.