Great strides have been made toward ease-of-use in today’s relational database systems. The query syntax and navigational requirements have been simplified. However, they still pose serious problems for the casual user who may not know important details such as available tables and their system names, or who may be seriously disadvantaged trying to formulate a complex query. Natural language, on the other hand, is often ambiguous. It does not automatically allow for system specific naming conventions, and it creates unrealistic expectations of the system’s power.
In this paper, the author introduces a database interface called HERCULES (HEuristic Retrieval: a Casual User LanguagE System) which integrates the formalism of relational query languages with the flexibility of natural languages. A natural language interface uses predefined information about the database to help the user interactively formulate a query. Once the proper query is constructed, it can be submitted to the database to be satisfied. The author developed a prototype in ULISP to demonstrate the feasibility and compared it favorably in several examples with Query Be Examples (QBE).
The author presents a lot of good ideas through this prototype; consequently, this paper should be of great interest to anyone interested in database query languages or the integration of database theory with artificial intelligence. There are two potential problems. First, it is no major conquest for a tailored, well-targeted prototype such as HERCULES to outperform a full-blown implementation such as QBE in certain specific instances. Second, the prototype, if implemented, would place a heavy, if not impossible, burden on the DBA to define the database fully to HERCULES since the data it requires is far beyond the data normally captured in the data dictionary.