Natural language interaction using a scalable reference dictionary
A truly natural language interface needs to be feasible for actual implementation. We developed such a new approach for database query and tested it successfully in a laboratory environment. The new result is based on metadata search, where the metadata grow in largely linear manner and the search is linguistics-free (allowing for grammatically incorrect and incomplete input). A new class of reference dictionary integrates four types of enterprise metadata: enterprise information models, database values, user-words, and query cases using an ontology-based meta-structure. The layered information models allow user-words to stay in original forms as users articulated them, as opposed to relying on permutations of individual words contained in the natural input. These properties make the approach scalable to the number of users and the size of the database. A graphical representation method turns the dictionary into searchable graphs representing all possible interpretations of the input. A branch-and-bound algorithm then identifies optimal interpretations, which lead to SQL implementation of the original queries. Query cases enhance the metadata and the search of metadata, as well as provide case-based reasoning to directly answer the queries. This design assures feasible solutions at the termination of the search - i.e., the results always contain the correct answer.
Full Text: PDF