An update expert and response generator for a transportable natural language interface to database management systems

Fully transportable natural language interfaces to database management systems (DBMS) have been under study for some years I but until now I all have suffered from a lack of response ability and lack of natural language update ability. Response generation is relatively easy to overcome, but the second problem, lack of update ability, is more serious. Adding update capacity involves primarily three tasks. First, the system must be able to recognize and process update requests. Processing an update typically involves both altering the knowledge base to reflect the new state of the database and performing dynamic extensions to the lexicon. Second, the intermediate language used to communicate with the database manager must be extended to cover update information. Third, the post-processor must be extended to transform commands into DBMS update requests.

The system described here uses a flexible and unified knowledge base to recognize and process update requests. Through information stored in the knowledge base, the system can recognize and resolve certain classes of ambiguity. The update request is then converted into an unambiguous intermediate query language. This language is easily translated to the target database management language using simple syntactic methods. The response generator uses the intermediate query language, the knowledge base I and the results returned by the target DBMS to form a response for all database accesses.