Understanding without proofs
Yorick Alexander Wilks · International Joint Conference on Artificial Intelligence · 1973
The paper describes the analysis part of a running analysis and generation program for natural language. The system is entirely oriented to matching meaningful patterns onto fragmented paragraph length input. Its core Is a choice system based on what I call semantic density. The system is contrasted with (1) syntax oriented linguistic approaches and (2) theorem proving approaches to the understanding problem. It is argued by means of examples that the present system is not only more workable, but more intuitively acceptable, at least as an understander for the purpose of translation, than deduction-based systems.