How can we reach a retrieval quality with a search engine that is comparable to a subject enquiry in a library catalog with a controlled vocabulary.

We evaluate different approaches for automatic document indexing by means of the SyynX technology with repect to an possible improvement of retrieval results.

At the end of the project, we implement an automatic document indexing and retrieval system that is used by the University Library. It is important for us that other libraries can use these results and the output of our system as well, without being forced to use the commercial SyynX technology. The generated metadata and the automatically assigned keywords will be available in a computer readable format that can easily be integrated in other systems.

We assume that an optimal result cannot always be obtained by automatic indexing. On the one hand, this can be caused by the employed method, but also by the nature of the used thesaurus: certain thematic areas often aren't covered sufficiently or not at all.

In connection with this, we examine, whether the last-named problem can be solved by the use of different thesauri at indexing step. As the economic sciences are an inherently interdisciplinary field of research, it has to be examined, whether the use of additional thesauri (particularly law and social sciences) from related domains can lead to better results.

On the basis of documents indexed with multiple thesauri, we try to create correspondences between these different thesauri that can be used to access contents directly that are indexed by only one thesaurus.

With this project we focus on the economic sciences and the indexing of economic journal articles published by Elsevier. This is especially desirable as the German Science Foundation purchased a national license for these journals and all German universities have access to these articles. So this project should improve the usability of this purchased license and contributes to the sustainability of the employed financial resources. As mentioned above, we give access to the automatically assigned keywords in a computer readable form, so that they can be used for example in ECONIS. So the results would be available for the virtual library for economics and business studies (Project EconBiz).