| Resumen |
The first Natural Language Interfaces to Databases were built and designed for specific domains, and their customization processes implied source code manipulation. Open systems and database inter-operability enabled these interfaces to be independent of the operating system and database management system, and the separation of the knowledge base from the translation process allowed for domain portability. Although commercial interfaces incorporate semi-automatic configuration wizards that help configure the interface without knowledge of its inner workings or its source code, it is still difficult to customize these interfaces for a given database, due to confusion on the information that is necessary to provide to the knowledge base of the interface in order to make it able to answer some query category. For solving this problem, we propose an ontology whose design is simple and flexible enough to assist the customizer's work. This paper describes the design of the ontology, as well as an empirical evaluation of this approach versus the customization process of a commercial interface. The evaluation was useful to detect problems with different types of queries used to retrieve information from a specific database. In spite of the difficulties found to make the evaluations and some unquestionable advantages offered by commercial interfaces. |
