CORDIS Archive

7.1 Data Model
7.2 Subject Indexing
7.3 Recommendations for the use of CERIF 2000 guidelines
7.4 Implementation Scenarios

7.1 Data Model

The major design objectives for the CERIF 2000 data model were:

• A full CRIS data model with flexibility to allow the majority of existing CRIS to accommodate their own database structures;
• A base framework for data exchange.

These objectives have been met by:

• Defining a full CRIS data model which will cover the database structures of the majority of existing CRIS;
• Defining a set of data models that could provide examples for data exchange (since there are an infinity of possible exchange data models between CRIS). These examples of data models also illustrate that it is not necessary to implement the full CRIS data model if the requirement is for only a particular subset;
• Defining a metadata data model to provide a uniform summary-level view over heterogeneous information sources.

The Ortelius Consortium made the following proposal to the Commission for the use of the Ortelius Thesaurus as the main R&D subject indexing scheme for CERIF 2000.

"The Ortelius Consortium agrees to undertake the following actions:

• To make the Ortelius Multilingual Thesaurus available via the CERIF Homepage (/cerif/) for browsing purposes.
• To distribute the Multilingual Thesaurus on a CD-ROM to authorised users. Authorised users will be identified by DG XIII-D, based upon their involvement in CERIF. The Thesaurus will have an international standard format, with export and import functions into any database for indexing and retrieval purposes. Technical specifications will be included, as well as routines for printing the Thesaurus. The CD-ROM will be updated and distributed every 4 months;
• To provide continuous management of the Thesaurus, according to the ISO guidelines for the establishment and development of a multilingual Thesaurus (ISO 5964). All terms or modifications proposed by the users through the CERIF user service via DG XIII-D will be considered for introduction into the hierarchical structure of the Thesaurus. The translation of the terms into the various languages will be a joint task.
• The Ortelius Consortium proposes that the Commission provides them with a financial contribution to support the costs of developing the CD-ROM, updating and distributing it, and to recover (partially) the costs of the staff involved in the management activities."

Contact person :

TOP

CERIF 2000 is intended to be "implementation independent". A discussion on implementation scenarios is included here to help the reader to visualise how it could be put into practical use.

Two extreme scenarios can be considered:

• The most simple case - Local implementation of CERIF 2000 to create a research information system;
• The most sophisticated case - Integrated implementation of CERIF 2000 for different sources accessible via a WWW-based Catalogue Metadata model, including rules for access rights etc..

To provide one single access point allowing the CRIS user access to distributed research information sources, one could consider a web implementation of the Catalogue Metadata model.

It is possible for the metadata and for the data to be in many different kinds of system. For general undisciplined web resources, it is becoming increasingly common for metadata to be in RDF/XML pointing to data in html. However, most of the data on the web is in databases, the html is ephemeral, and the metadata is also stored in databases - in the form of data elements that could be converted to XML if required; - relational systems are particularly suitable for this. This means the metadata and data should be modelled correctly and then implemented in the most suitable systems environment. The relative strengths and weaknesses of the different types of implementation system environments (Relational, Object Oriented (OO), Information Retrieval (IR) and RDF/XML) are discussed in the following section. Then the practicalities of metadata extraction are reviewed before a discussion on how it might be implemented in ERGO, in RDF/XML and in an IR environment.

Schema metadata has to be constructed by the database administrator, although in advanced heterogeneous systems it is possible by schema reconciliation to generate new schemas. A similar situation applies for metadata for security, access or charging purposes.

Content metadata is different; instances are subsets of the real data - instance by instance - a catalogue. This can be generated by projection from the data instances - but this requires the content metadata to be a proper subset of the export or exchange schema (what the Database system is willing to expose to the outside world). This is the basis for the three layer model of CERIF 2000.