Conditional answers in question-answering systems
Philip Klahr · International Joint Conference on Artificial Intelligence · 1979
In many cases a deductive question-answering system cannot find complete proofs to answer questions requiring deductive support. In some cases information needed to complete proofs is missing from the knowledge base. In other cases processing limits may have been reached before proofs could be completed. Rather than disregarding such partial proofs as most systems do, the DADM system displays them to users and identifies subproblems that remain unresolved. Answers emanating from partial proofs include remaining subproblems as conditions which must be true for the answers to be valid.