Dividing up the question answering process
Marc Luria · National Conference on Artificial Intelligence · 1982
This paper describes a question answering program which divides question answering into two separate processes: answer formation and answer expression. Rather than gathering possible answers and choosing the best among them, the program accesses the database and finds all components of possible answers, e.g. a causal chain, and then passes this information to an expression program which formulates a proper answer.