Natural Language Interfaces to Database Systems.

摘要

Database management resulted from a need for data to be retained in the computer beyond the period of the current run. Since the earliest times of computer interaction, retrieving this data has been a significant problem. The motivation for this study was to look at the history of database in order to critique the human computer interface with databases and to project the next areas of research in database management. A chronic consequence of man's normal condition of possessing incomplete knowledge of the detailed information to access and use that data is query failure. Although all query failure is frustrating, natural language failure can be especially insidious because it can fail by giving erroneous information to an unsuspecting user. Despite the current level of enthusiasm for natural language communication with the machine, this emulation of human-human communication may be the wrong approach to improving human-machine communication. A better approach may be to use methods of artificial intelligence such as semantic networks and object-oriented programming to create information sublanguages. Such a sublanguage would relieve the user of the need to have knowledge of database meta-data. Adaptive methods, flexible interaction, conceptual pattern matching and disambiguation methods paradigms that can be employed into a new (i.e. beyond relational) hybrid database to strengthen the underlying structure of databases before communication with them can be improved. This paper provides suggestions on research directions in order to achieve that strengthened database structure. Keywords: Natural language interfaces; Man computer interfaces; Database interfaces; Artificial intelligence.(kt)