ASAP to REL: Efficient Relational Data Bases from Very Large Files.

摘要

Recently, several very powerful interactive question answering systems have become available. In general, these systems permit the user to pose questions in English or some formal language and will perform searches through a data base which is structured as a set of relations. These searches may involve deductive or inferential logic ability. There are two major difficulties which inhibit the widespread use of such systems. First, these systems typically deal only with a 'small' number of relations, e. g., less than 1,000,000. The data bases which we wish to query in real life are usually much larger. Second, the creation of the relational data base itself is often quite time consuming. In this paper, the problem is discussed of taking a very large data base stored in conventional rectangular form, i.e., all data elements dealing with one entity (name, address, phone, balance, for example) which are stored contiguously, selecting a subset of interest, and reformatting it into a relational data base which can be used by a question answering system.