Next-generation library catalogues: review of E-LIB Bremen

The electronic library of the State and University Library Bremen (E-LIB) is based on search engine technology and offers an integrated catalogue of electronic and print media. All of the library’s available media and services can be accessed via a single search interface (based on the idea of a ‘one- stop-shop portal’) instead of the separate catalogues for different media types still to be found in many libraries. The idea behind this system is to combine all kinds of local holdings and all the external resources available for our users in one single retrieval system and to offer new methods for supporting associative search processes.

The electronic library of the State and University Library Bremen (E-LIB) is based on search engine technology and offers an integrated catalogue of electronic and print media. All of the library's available media and services can be accessed via a single search interface (based on the idea of a 'onestop-shop portal') instead of the separate catalogues for different media types still to be found in many libraries. The idea behind this system is to combine all kinds of local holdings and all the external resources available for our users in one single retrieval system and to offer new methods for supporting associative search processes.
There are about 20 million records -nearly 80% of these are electronic full-text documents -in the library-controlled web space of the E-LIB Bremen. The technology behind this system is based on a locally hosted index of heterogeneous, aggregated metadata. The data is organized by the CiXbase 1 system that is not only used and refined in Bremen but also at the ETH Zürich (NEBIS) in Switzerland. The E-LIB Bremen system is both a live and a test system for new services 2 .
Typical advantages of a search engine using a local index are the high response rate, the homogenous display of results 3 and a well-balanced result list. In addition, the metadata of search results establish a knowledge base that offers additional options for new forms of analysis that can be processed during the search.

Local indexing and interoperability
Various types and formats of metadata are imported and converted to a homogenous XML flat file structure in order to build the local knowledge base. The resulting consistent metadata pool is indexed via search engine technology based on statistical balanced B*-trees algorithms serving as a basis for the search engine. The results are presented in form of XML-structured metadata within the browser. This XML data is converted by stylesheet-based XSL transformation within the client's browser application 4 . This separation of structure and layout enables nearly complete interoperability of the available metadata and simplifies further processing of this data in other information systems as well as the integration of external services into the E-LIB.
Export functions for citations are provided mainly for different bibliographic database formats such as RIS, Endnote, etc. The 'import into reference database' software is particularly convenient if the data can be delivered to a webbased system like RefWorks or Connotea. This open display architecture has also enabled us to extend our result presentation easily via an integration of web services of other external information providers -thus realizing mashup functions.

MARTIN BLENKLE, RACHEL ELLIS AND ELMAR HAAKE
Additional information related to the book titles is integrated using an experimental service from Germany's Common Library Network (GBV) 5 called SeeAlso. For example, we integrate links to titles in Google Book Search, LibraryThing or Wikipedia in the record presentation, if available from the service provider.

Discovery search and recommendation features
The search box is always present at the top of the screen throughout the whole session -a user can start a new search at any point. The results are presented in a two-column display. The left column displays the results, while the column on the right provides the drill-down options and additional features (see Figure 1).
A search within such a large database often results in a confusing number of results. The search process thus has to be supported by adequate methods of refining the search inputs. In addition to formal and content-driven drill-down functions, you find a number of recommendation tools for the modification of search results. During the search a user has the chance to alter the search strategy in the sense of 'discovery search'. The subject-based search modus then turns into a multilevel iterative navigation within a semantic net that has been built on the entity of all available subject-indexing data of the bibliographic records.
In order to achieve this, all the material of controlled vocabulary of search results is analyzed, statistically evaluated and presented as a tag cloud (see Figure 2). The result of this analysis -the relevance profile of the tag cloud -depends on the chosen term as well as the available title material and its subject index elements.

Handling of 'no hits'
The analysis and adequate handling of 'no hits' results is an interesting application for a search engine assistant. No or just a few hits are usually either caused by titles actually not to be found in the knowledge base or by spelling mistakes. The E-LIB search engine therefore offers a spell checking and recommendation ("Did you mean") facility -using the principle of the Levenshtein-Distance algorithm that works with various languages due to the index material available.
In the case of no or just a few hits, the E-LIB system checks the availability of data in the common library network's union catalogue via SRU request. In case of a successful response, an order button is offered without switching to another system. This method can also be used to combine hits from external bibliographic databases with the local holdings -always considering the access conditions for those as well. In the near future the intelligent combination of local indexing and cross-searching will combine the major advantages of both search techniques.

Service-oriented architecture
In addition to the classic search functionalities, one aim of the E-LIB system is to offer new views of the available metadata collection in order to support the data management on the campus in a convenient way. Our students can use the whole variety of E-LIB's knowledge base directly without diversion within their learning management systems via SRU interface. The results are presented in the 'look-and-feel' of the course system. As the search engine offers access to more than 17,000,000 electronic resources with full-text links -among those thousands of e-books -these can easily be used to create electronic course packs.
In addition to the search engine, many of the other E-LIB services are designed in a modular way and can thus be integrated as an independent service into external systems. Such services, e.g. the tag cloud generation and the subject analysis of the request as basis for database recommendations, can be supplied. Provided that the knowledge base has a certain size, is comprehensive and evenly spread on the different subjects, it can be assumed that the services can reach a certain degree of independence from the local holdings and can universally be integrated into other systems.