Focus of this thesis is the relationship between databases and information retrieval systems. As a background, the first part consists of a general presentation of databases and information retrieval systems and some examples of already existing efforts to combine the two. While these examples typically have expanded either a database system or an IRS to obtain multi-functionality, we have made an effort of bridging the two systems.
Our prototype integrates FDS (Fast Data Search) into the PostgreSQL database management system as a new index access method. FDS is a powerful and scalable commercial enterprise search platform using a typical search engine query language. PostgreSQL, being open source and a general basis for research, lends itself well to customization. The new index access method provides the database with powerful free text capabilities while retaining the power of the relational model for structured data. Preliminary results including a simple performance test verify the feasibility of the integration, and demonstrate the scalability of the prototype. Storage, indexing, updating and search functions are implemented, but ACID properties could not be guaranteed, because the external indexing system has no such guarantee.
I also present a prototype for automatic extraction of related structured data in the relational database to XML. Combining these two prototypes by allowing the extracted information to be searched using the full text index, makes it possible to search the database without knowledge of the underlying database scheme.
Finally I discuss potential expansions of our implementation by indexing other data than text, multicolumn-indexing and moving complex evaluation from PostgreSQL to FDS, and suggest how this could be done.
The thesis is written in Norwegian.