MACS Project Enters Third Life

Around the change of the millennium, a few National Libraries united in CENL decided to embark on an ambitious project: to semantically link up their native Subject Heading Languages. The libraries felt a growing need to offer end-users a reliable and high-quality crossover from, say, a French keyword system to a German one. This project became known as MACS: Multilingual ACcess to Subjects.

We write the year 2000. As information science researcher at Tilburg University, I got asked by the Library of the university to assist in writing a tender from CENL for a prototype management system for these subject heading crosslinks. Together with Index Data from Denmark, we submitted a tender -- and won it, over eleven other European companies and institutions.

It turned out that it was mostly the proposed design of our system that won the race. We did not propose to re-use an existing system to kludge links between existing catalogs, but to develop a simple though well-tuned classical database application as link management front-end. It was complemented by a classical XML-based search engine, Index Data's Zebra, to demonstrate the multilingual search capabilities over a 200,000-volume library database sample. The system got successfully implemented, and this led to CENL awarding us the contract for the production version.

This production version was a rebuilt, ruggedized version with extensive workflow capabilities. We designed it to handle a million links between up to fifty subject heading languages, though we knew that these limits would not be reached within a decade. The system was built fully with OpenSource software and went into production around 2005.

Especially the colleagues at the Swiss National Library, first led by Genevieve Clavel and later by Patrice Landry, were the force behind MACS. Having a multilingual library to run, they were used to having several independent subject heading languages at their hands. Their team of link editors managed to extend the MACS database to close to 100,000 link records, largely by manual effort. Although not all links are trilingual (French RAMEAU, English LCSH, and German SWD), the Swiss people aim squarely at a full cover of all SWD headings they use in their own index.

In 2008, some improvements were made to the link management interface, and experimental export of the semantic linking data to a new search interface was begun. This led to a prototype MACS-supported interface on top of The European Library. While the interface is not yet ready for the European public, first tests are very promising, retrieving documents cross-language out of the catalogues of four National Libraries. Truly a European first!

For 2009, we have set up a new machine for MACS and the TEL interface, and we are currently organising a few significant work flow changes in the system. These changes will accommodate the move from link creation to authority maintenance. The teams become larger and geographically more dispersed, and this requires better work division management. Additionally we integrate several Web 2.0-inspired features to the link management interface, top up the quality assurance, and integrate tightly with the available online authority databases.

All these efforts will lead to an improved, more efficient work process to get the linking done. It still is an intellectual effort -- no high-tech semantic technology goes into this project. But this may change. Now we have a well-maintained, professionally staffed database system that lives, we invite researchers and professionals to deploy their technology and show off what it can do. The MACS database will be available to interested parties (under some restrictions), and we all know that with such a magnificent bootstrap, machine learning and other technologies will perform much better.

Stay tuned for more news about MACS! You can visit the MACS community platform at http://macs.cenl.org/.