A survey of question answering in natural language processing

Abstract This paper surveys the different basic approaches of Question Answering (QA) used over the last two decades. Our goal is not to give an exhaustive overview of QA-systems, but to capture the main trends and developments, the progress being made, and some of the problems still remaining. QA-systems are classified as procedural QA-systems, conceptual QA-systems, and logic-based QA-systems. Procedural QA-systems belong to the earliest attempts to model question answering by attaching specific procedures to words. Conceptual systems emphasize the aspect of meaning representation of questions, memory, and answers. Logic-based systems use logic predicates to represent knowledge for question answering. General issues of noun phrase disambiguation, variability, and cooperative behavior are described as examples of problems during question analysis, answer generation, and user modeling. Uniform knowledge representations, logic representations, dictionaries, and transportability are identified as current trends in question answering.