Manjula Patel
UKOLN
University of Bath
Claverton Down, Bath, BANES BA2 7AY
United Kingdom

Robina Clayphan
National Bibliographic Service
The British Library
Boston Spa, Wetherby, West Yorkshire, LS23 7BQ
United Kingdom

Introduction

National bibliographic agencies (NBAs) across Europe differ in their selection policies, which are typically based on the prevailing legal or voluntary deposit arrangements. Currently however, they are all facing similar challenges with regard to the changing practices in publishing brought about by developments in electronic publication. Until a few years ago material was published by a relatively limited number of organisations in any one country, most of whom had well established lines of communication with their national library and were well aware of the legal and commercial framework in which they operated. Given the explosion in electronic publication, it has been envisaged that significant numbers of valuable publications could go unrecorded since they would by-pass established national bibliographic procedures.

The need to research and develop systems to address the ensuing issues was recognised by CoBRA (Computerised Bibliographic Actions)[1]. A proposal was therefore put forward to develop a system which would allow publishers to provide an agreed minimum level of data to NBAs. The data could then be enhanced to national library standards and transmitted back to the publishers for inclusion in the publication or within their promotional literature. The purpose of the BIBLINK project[2] is to create a prototype system that will provide an automated flow of metadata between the two sectors.

The benefits for the national libraries is clear, an early receipt of bibliographic information in machine readable form would provide for editing and entry into the national bibliography with the attendant possibilities of developing new products and services relating to electronic publications. The pincipal benefit for publishers would be the recording of the publication in the national bibliography with the concomitant prospect of wider public awareness of its existence.

To carry out the work of the project five European national libraries (UK, France, the Netherlands, Spain and Norway) joined with two academic institutions (UKOLN at the University of Bath and the Open University of Catalonia) and two publishing trade organisations (Book Industry Communication and CD-ROM Standards and Practices Action Group) to form the BIBLINK Consortium.

Objectives

The BIBLINK project was proposed with the objective of developing a demonstration system for acquiring metadata for electronic publications directly from publishers. Not only does this enable electronic documents to be incorporated into the national bibliography, but it also allows for the resulting data to be returned to the publisher so that it can be embedded in the publication itself, or be used in other applications, such as sales promotions. In addition the enhancement of the author data, by means of authority control, and the addition of authoritative subject data could potentially enhance customer searches of their own catalogues and other listings should publishers choose to use the returned bibliographic metadata. The overall objectives of the project can be summarised as follows:

Project Structure

The project officially started on 15 May 1996 and was divided into two distinct phases, each of which was expected to be eighteen months in duration. The first phase was primarily for research and consensus building and the second for software development, implementation and testing.

During the first stage, in addition to the scope of the project being more precisely defined, information was gathered and analysed about the various components that contribute to a data generation, transmission and conversion system. This included detailed studies relating to data formats, numbering systems, encryption, authentication and format conversions. The research entailed considerable discussion with publishers and consensus with them as to the most promising technical solutions to implement. Further information may be found on the project's web-site[2].

In the second phase of the project, work began by formalising the user requirements and producing a detailed functional specification[3] which has resulted in a prototype system[4]. An exploitation plan is being developed to provide a framework for library partners to assess the possibility of incorporating the system into their operational procedures. At the time of writing, the project is in the Demonstrator phase, during which use of the system will be verified by several national libraries and over twenty publishers.

The BIBLINK CORE Metadata Set

A metadata study[5] undertaken during stage one, identified current approaches to encoding metadata. A comparative analysis was carried out and recommendations made as to which were most likely to be of use in the project. Given the range of publishers involved it was felt that offering only one format would not be realistic. The recommendation was to use the element set defined by the Dublin Core Metadata Initiative (DCMI)[6], extended as appropriate to accommodate the minimum requirements specified by the libraries. These were mapped to the MARC formats in use in libraries to ensure the data provided would produce a sufficiently rich record.

The element set adopted is shown in the table below. It comprises twelve of the fifteen Dublin Core (DC) elements plus another ten defined by the partners as the minimum data set libraries would require. Those shown with a DC prefix in the table are the elements taken from the Dublin Core, and those with a BIBLINK prefix are the ones added by the project.

Table: BIBLINK Core Metadata Set

The elements of the BIBLINK Core (BC) were mapped to UNIMARC as an intermediate format. Since an SGML format used by more than one publisher could not be found, Chadwyck-Healey’s DTD was also mapped to UNIMARC to prove the feasibility of using SGML. USEMARCON, an existing conversion tool developed by another EU project[7], was extended to accept and process the incoming data. The output formats needed were the MARC formats of the participating national libraries (UKMARC, IberMARC, InterMARC, BibsysMARC) with the exception of the Netherlands whose MARC format does not comply with the ISO2709 standard generally used for MARC records. In addition a DC tagged HTML view was needed so that records could be cut and pasted into the source code of a web publication. The system also generates a simple listing of the element labels and their values.

Identifiers

A study of identifiers[8] resulted in a list of requirements to cover the range of resources included in the scope of the project. The chosen identifiers had to cover both on-line and physical items as well as exhibiting the characteristics of persistence, global uniqueness, extensibility, human-readability, and of being capable of transported over the internet. Ideally they would be standard and come from an authorised naming authority.

No one identifier was found that would meet all these requirements. It was recognised that it was beyond the scope of the project to enforce the use of any particular scheme in any case and that the project should offer a range of options. Those actually incorporated into the project include the established schemes of SICI, ISBN, and ISSN, and those currently under development for digital items: URN and DOI. For all its shortcomings as an identifier the URL was also included as it would be unrealistic to exclude the scheme that was most likely to be offered by publishers.

The BIBLINK Workspace

An overview of the system produced is shown in the figure below. The BIBLINK Workspace (BW) can be characterised as a computer-mediated shared workspace for the creation, processing and transmission of metadata. At its core lie the database of records, the conversion software and the system control software. These communicate with email and Web interfaces for the receipt and automatic onward transmission of metadata.

There is a two-way channel of communication for publishers to submit data and receive enhanced data in return, similarly, an interface for NBAs to receive publisher data and return it. If identification agencies are to be involved they can be given the right to read the record and add data to the identifier field only. Access to the BW is governed by a user name and password allocated by the BW Administrator.

Each channel of communication is available using either an email or Web interface. Publishers can elect to use whichever they feel is most appropriate to in order to submit their bibliographic data. The Web interface was intended for those who have no systematic bibliographic database from which to extract data. It provides a straightforward method for entering text and the option of nominating a scheme for identifiers if appropriate.

The email route is also simple but perhaps less straightforward. There are two ways of using it: an email message can be constructed using a syntax specified for BIBLINK with the data in the body of the text. Alternatively an existing SGML or HTML dataset can be sent as an email attachment. Either way the process could be semi-automated by writing a simple script to harvest the data from a database in the prescribed format and using an email template for the commands.

Figure: Overview of the BIBLINK Workspace