An advanced new reading station for the Bibliothèque de France

Under testing by a team of humanities researchers are two new prototypes of what may be the computer-aided reading station of the future for France’s library-to-be. An inside look at work in progress.

If everything goes according to plan, when the doors of the new Bibliothèque de France (BdF) open sometime in 1995, humanities scholars and other text-oriented researchers will be able to pursue their work at a new computer-assisted reading station (known by the French acronym PLAO), custom-designed to meet their needs.
    The DECision to develop a new PLAO was taken by the BdF Technical Committee in 1989 in the context of the library’s overall information technology policy. This entailed widespread digitization of holdings (100,000 works to be scanned into the book-base by 1995, 10% as text, the rest as graphic images) and the full computerization of the catalog, with an additional and as yet unspecified “linguistic assistance to catalog consultation” service under study.
    Elaborated under the leadership of Bernard Seigler, a‘reading’ expert at the University of Compiègne, the PLAO scenario specified that a networked workstation should meet the requirements of “professional” readers, working over a long time-span, requiring the individual selection of documents, and necessarily demanding a close relationship between reading and writing. Its implementation entailed the design of a high- quality user environment connected to the library’s own information network and made available for scholars in some 300 special study compartments. According to Bruno Forni of the Information Systems Design department of the BdF, the technical demands of this project make it “unlikely that this many will be available on opening.” The availability of digitized materials may also be less than hoped for, given the high error rate of current ocr equip-ment. Nevertheless, the PLAO project is now well underway and promises to provide a reference for future work in the field. The first phase investigated whether off-the-shelf software was a suitable source of PLAO tools for the humanities researcher. A team of seven scholars spent several months working with a variety of Macintosh software, including file navigation, text editing, and indexing software such as Mark-Up, Hypercard, Phraséa, Marco-Polo. Drawing on their comments, document architecture specialist Jacques Virbel of the University of Paul Sabatier, Toulouse, delivered a report to the BdF June last year, and his ideas were to form the basis for the technical specifications of the new PLAO.
    The main problem, Virbel felt, was how to shift away from the essentially writing context of current text editing interfaces or the reading models enshrined in electronic books or hypertexts and work towards the design of a researcher environment in which the prime concern was the combined act of reading and writing, i.e., accessing and reading text sources, adding personal notes to existing texts, and later navigating through them, using them in the writing process in a continually interactive way. Not surprisingly, therefore, the core concept of the PLAO was to become “the personalization of documents.” This meant that the PLAO should provide the minimal amount of pre-processing to researchers’ documents. “Ideally a bitmapped and a full-text version of everything,” recommends Bernard Quémada, the recently-retired director of the Trésor de la Langue Fran‡aise dictionary project, which supplies digital stock to the BdF. Researchers should further be equipped with a maximum number of on-line tools to build their own ‘document workspace.’.
    The next step involved actually designing a new PLAO. In June 1991, the BdF launched a call for tenders for a prototype PLAO, and of the ten replies, two were selected for the construction of a mock-up. In June of this year, the two prototypes were delivered for a second round of testing by the seven-member team of researchers. One model – built around the concept of the “personal corpus” – was designed by a team from Cap Gemini Sogeti (See  sidebar ), an R&D group with a strong interest in NLP and man-machine interfaces. The other was developed by Berger-Levrault ais., a specialist document-processing firm, which proposed an SGML environment synchronizing both graphic and text modes for the same document and enabling annotations with different properties to be affixed to texts, plus a set of tools for navigating through them. This environment does not call for inverse files, dictionaries, or other word-based resources. Nor were lexical or linguistic editing tools integrated into the model. On the basis of evaluations of these two approaches, BdF will draw up a set of final specifications for the production of a PLAO via a call for tenders in spring 1993.
    The ultimate PLAO scenario could well extend beyond the document-handling and on-line tools environment that the two prototypes address. According to Bruno Forni, “the technical committees concerned have agreed that the PLAO architecture should be designed to support further resources.” These could range from access to the growing availability of CD-ROM text corpora to linguistic facilities for text analysis, such as lexical extraction and even parsers, once these tools become sufficiently robust.