A data model, knowledge base, and natural language processing for sharing a large statistical database

Most existing statistical databases are mere collections of statistical files gathered for specific purposes. Consequently, as they grow in size, users are faced with difficulties in identifying and finding the data they need. In order to obtain data descriptions independent of specific purposes, this paper proposes an object-oriented data design, which distinguishes between data conceptually obtainable and data actually stored in a database, and specifies relationships among classifications and categories independent of particular data files. This is followed by a discussion of the representation of knowledge about data and classifications on a knowledge base, giving clear definitions of hierarchies and relationships among statistical data concepts. Finally, a natural language query system using the knowledge base is demonstrated, which proves the advantage of the proposed statistical data concepts.