Data Structures for an Integrated Data Base Management and Information Retrieval System

New applications like office information systems need interfaces to data bases which integrate classical data manipulation with management and retrieval of textual (“unformatted”) data. The relational data model is widely accepted as a high level interface to classical (“formatted”) data management. It turns out, however, to be inconvenient for handling even simple data structures as commonly used in information retrieval systems. To attack this shortcoming we propose an extension of the relational model by allowing Non First Normal Form (NF2> relations. We summarize extensions of the relational algebra, with main emphasis on the new “nest” and “unnest” operations which transform between first normal form relations and the NF2 ones. A related language is discussed on the basis of a hypothetical SEQUEL language. As a contribution to the roblem of efficiently supporting the access to NF g tables a novel index support for structured textual data is proposed. It is based on word fragments, words, and word sequences as internal (textual) keys.