METADATA SEARCH: A NEW APPROACH TO NATURAL LANGUAGE DATABASE INTERFACES

摘要

Metadata search is a new approach to process text natural language database queries. Its objective is to support any style of text inputs. To accomplish the objective, it builds a metadata reference dictionary containing information models, database values, user-words, and query cases. The reference dictionary accommodates any possible interpretations of a natural language query concerning enterprise databases. The branch-and-bound search method makes it possible and efficient to search all possible interpretations of a natural language query in order to determine the best solution. The approach also incorporates a case-based learning and case-based reasoning to assure successful closure to a query and to improve performance. A laboratory testing of a prototype shows the feasibility of the approach.