Investigations of Human Question Answering

摘要

This project developed and tested a model of human question answering (calledQUEST). QUEST accounts for the answers that adults produce when they answer different categories of open-class questions, such as why, how, when, and what-if. QUEST identifies the information sources for questions and assumes that knowledge is organized in the form of conceptual graph structures containing statement nodes and relational arcs. Example types of structures include goal hierarchies, casual networks, taxonomic hierarchies, and spatial partonomies. Question answering procedures operate systematically on these knowledge structures. An important property of QUEST consists of three convergence mechanisms which narrow down the node space from dozens/hundreds of nodes to a handful of nodes that serve as good answers to a question. Question answering, Text, Natural language comprehension, Computational models, Artificial intelligence. (eg)