IRCAM - Centre Pompidou Serveur © IRCAM - CENTRE POMPIDOU 1996-2005.
Tous droits réservés pour tous pays. All rights reserved.

The Multimedia Library at the Digital Crossroads

Michael Fingerhut
Bibliocom 2001, Rome, octobre 2001.
Copyright © Ircam - Centre Pompidou 2001

I would like to thank the Italian Library Association and Mrs. Andrea Paoli, Stefania Fabri and Palmira Barbini for having invited me to speak here before you.

New technology, old problems?

The impact of digital technologies both for storage and transmission (via computer networking) on traditional libraries brings to the fore many apparent dichotomies:

·        Physical place vs. virtual space

·        Preservation vs. circulation (or archives vs. public collections)

·        Bibliographic records vs. metadata

·        Knowledge vs. information


So-called “traditional” libraries used to have only paper-based collections; even then, they were constituted of more than “mere” books, to encompass other types of objects such as musical scores, maps, or engravings, say. Through the centuries, they have developed rigorous – some would say rigid – systems of classification, search and retrieval for these real, physical objects, themselves based on paper. Both the objects and their classification schemes were directly accessible by humans without any need for technological means.

With the appearance of electromechanical means of recording and production, libraries were faced with a multiplicity of media – whence one of the uses of the term multimedia – used to document the process of human creativity, recordings, movies, CD-ROMs, databases and hypermedia objects… These, in turn, differ from books in several ways:

·        They usually need some technological means of access – audio player, VCR, computer;..

·        They may be “somewhere else”, outside of the control (but not of reach, at least for a time) of the library.


It is no surprise that libraries have traditionally separated paper from other media. In some major libraries (I won’t mention names, here) musical scores and the recordings of the same musical works are to be found in separate departments, in different buildings. Yet, how “natural” is it to listen to a score and read it at the same time?

Another historical separation exists between archival material – e.g., recordings of concerts, conferences and other events – and “library” material, even though both may consist of almost identical objects, such as the recording of a particular musical work, the only difference being that the “archive” had not been made commercially accessible.

It is then this major challenge that libraries have to face: how to integrate this variety of objects, not only as a mere juxtaposition in the same place, but also how to provide the intellectual coherence and ease of access.

However, this is not really altogether new: multidisciplinary environments have always had to face the juxtaposition of different disciplines, systems of thought and modes of transmission and address the necessity to provide a coherent intellectual view. In the remaining of the talk I will address these issues, with the specific point of view of my organization.

The Ircam Multimedia Library and its Multidisciplinary Environment

The Ircam Multimedia Library (term which we use to translate the French word médiathèque) is indeed a case in point. IRCAM[1] is a not-for-profit organization, founded in the late 70s by composer and conductor Pierre Boulez, as a multidisciplinary research institute dedicated to contemporary music. Its personnel consists of composers, musicologists, choreographers, performers (instrumentalists, dancers), researchers (computer scientists, acousticians, specialists of perception, of digital signal processing, sound engineers), lecturers and students (of composition, musicology, computer music)… all with a unifying goal in mind – contemporary music creation, from its varied, almost infinitely different, perspectives: technology, science, art, philosophy, sociology... Ircam is associated with the Pompidou Center, which is the French national center for arts and culture, itself  a multidisciplinary institution in the artistic domain: painting, sculpture, cinema, music, industrial art… This adds another dimension to the tensions I mentioned above, which has an impact on the library of such institutions:

·        Tradition vs. modernism (or preservation vs. creation)


The IRCAM Multimedia Library was created in the mid 90s based on the (traditional) collections of the music library then in existence at IRCAM, with the addition of different archives, collections and technology. It is, at the same time, a specialized lending library and documentation center for the personnel of the institution, and a library open to the public. In addition, it provides such services as various international mailing lists[2], technology newswatch, collaboration in projects related to digital libraries and metadata, as well as expertise and counseling in these and related domains.

Its collections comprise the whole gamut of media we mentioned:

·        Books, paper and electronic periodicals and scores, in print or manuscript, of  contemporary music

·        Archives of IRCAM concerts (recordings, program notes)

·        Digitized scientific and musical publications of IRCAM

·        Commercial recordings of contemporary music

·        Documentary films on contemporary music and dance

·        Databases on contemporary composers and their works

·        CD-ROMs (an endangered species which will probably become extinct in a while)

·        Multimedia objects (music analyses, score and performance objects…)

·        Internet access


The major challenge was to provide a unified view of these collections, together with a simple and reliable means of access to the objects they contain. In order to do so, we followed a basic principle:

·        All objects are (born) equal


Zone de Texte:  
Figure 1. The audio interface, showing the structure of musical works

For instance, this means, in our case, that the recording of a piece found in an Ircam archive “looks” the same to the library patron as a commercially-available recording, both from the point of view of search (catalog) and of retrieval (interface).


The design of adequate metadata for our specialized, multimedia environment had to take into account the search strategies we wanted to provide to our users. In the case of musical works, several families of criteria can be proposed to the patron:

·        Inherent to the piece (composer, title, performers, places and dates, topic, dedication,etc.).

·        Related to the piece (scores, other like pieces – by period, genre, style, composer, topic –, related articles – about the piece, about the composer, about a performance…)

·        By contents (instrumentation, genre, style, theme, motif, rhythm, citations, lyrics…)

·        Audio and psychoacoustical (e.g., timbre, mood, etc.)

·        Fuzzy (“I heard yesterday a bebop-like piece sung by a man, which spoke of…”)


Since the Multimedia Library emerged, as if it were, from a library, the choice was made then to start with traditional bibliographic means of reference – UNIMARC records, which are the product of a long tradition of cataloguing and indexation. Yet, since we wanted to address the above list of issues, we had to adapt our cataloguing strategy in various ways – either extension (for instance, in order to catalog musical scores with detailed instrumentation), use of alternate formats (for instance, for music recordings, in order to show the structure of the piece, see figure 1), of distinct databases (for instance, for biographical information on composers and musicological documentation on their works), thereby satisfactorily answering the first two concerns, and part of the third one. But while the interface (see below) masks this variety of underlying mechanisms that had to be put in place, this multiplicity and sometime redundancy  is a shortcoming of the technology in existence at the time when the system was designed.

This is where the issue of bibliographic records vs. metadata arises. While both intellectually address the same need – describe intellectual production –, they emerge from radically different fields, the former in order to describe physical objects and their localization in a library, the latter in order to reference and access objects on the Internet. This is why they currently differ (see the table below), but the inescapable rapprochement between these two domains provide for mutual influence.

From our point of view, there is a shortcoming in both systems in that they still tend to address flat objects, be they physical or electronic. We need to be able to reference higher-level and hierarchical intellectual entities, and this is what the so-called FRBR[3] and its extensions wish to address. Such models (see figure 2) are composed of entities and relationships, and may express the whole gamut from the abstract notion of work to that of a physical copy, e.g.:

·        Zone de Texte: 	Computerized bibliographic records	Metadata
Standards	MARC (UNIMARC, INTERMARC, UKMARC, LCMARC, MARC21…), MABStable, controlled evolution	Dublin Core, RDF, MPEG-7…Evolutionary
Record localization	Independent of the document (may include links to electronic documents)	Associated to the document, included in the document, including the document, independent of the document
Number of fields	Fixed and hypertrophied (UNIMARC: 10 blocks of 100 fields, with many subfields)	Limited (Dublin Core: 15, with possible extensions via qualifiers), or open.
Contents control	Well-developed: typed, authority control, thesaurus…	Under-developed.Recent developments: relations
Record production	By professionals	By the author, automatic, etc.
Authentification	Yes	Sometimes
Intellectual property	No	Yes
Format	Rigid, difficult for hierarchical data	Extensible
Concepts vs. objects	See FRBR and DC extensions

[w] J.-S. Bach: Six suites for cello solo

·         [e1] Manuscript score copied by Anna-Magdalena Bach

·        [m1] Facsimile publication by Bärenreiter

·        [e2] Interpretation by Janos Starker (1963/5)

·        [m1] Recorded on a 33rpm vinyl by Mercury (1965)

·        [m2] Reedited CD by Mercury (1965)

·        [i1] Copy on a shelf…

·        [e3] Interpretation by Pablo Casals


Zone de Texte:  

It remains to be seen, however, if they can also be used to describe the hierarchical contents of a work (acts, scenes and movements in an opera, say) and how they relate to the corresponding segmented parts of an item (tracks on a multi-volume CD, say).

Another good reason to sit on one’s hands is the fact that new standards, such as MPEG-7 or Dublin Core, are in evolution, and have not produced yet reliable and long-lasting, stable collection of metadata, as is the case for the various MARC systems. It is to be hoped they will not produce hypertrophied “cataloguing” systems so complicated that they won’t be of a real and practical use.

It is sometimes argued that in the face of complex metadata, there is no need to catalog anymore, but merely obtain the metadata elsewhere, or even just reference it, by means of an identification code. While the idea seems appealing, there is a plethora of such codes and systems[4], with sometime overlapping domains and unclear relationship to the FRBR model.

In the current situation, our short-term plans are to extend our current library systems to include other metadata alongside the UNIMARC records within the same software system, thereby sharing the controls, indexes and thesauri, and allowing for the specific referencing of radically different objects and concepts.

Indexation is another issue at stake in such a multidisciplinary and specialized environment: the available mechanisms (classification systems, thesauri, etc.), have a wider and shallower coverage than the one we need. We have thus been led to develop our own specialized classification schemes: for the musical genres (see figure 3), for the monographies. They allow for much more precise searches of material in our collections than would be provided by traditional systems.


From the inception of the project in 1995, two key decisions were made as to the interface:

·        Every “non-print object” (e.g., movie, sound archive, audio CD, etc.) had to be made exclusively available by electronic means, with adequate protection against illegal access and pilfering.

·        Zone de Texte:  
Figure 3: indexation of musical genres

The Web (i.e., HTML) would be used as the unified interface to all the digitally available material, on a single terminal. This interface would use familiar representations (e.g., resembling physical objects such as audio players, VRCs and the like), be simplified (all “unnecessary” controls and menus are suppressed) and secured as well.


As the consequence of the use of hypertext (and hypermedia), the lack of a single, global, information system as mentioned above is masked to the user. Every “object” is linked to other intellectually related objects via hyperlinks (see figure 4, in which the bibliographic record is linked to biographical records in a different database as well as to electronic audio contents).

This, in turn, provides two major benefits:

1.      The ability to show different views and perspectives of the same underlying collections.
In this way, the reader can search either for all documents (of whichever form – book, film, recording, etc.), or search in particular collections (say, online articles). He can even use a three-dimensional interface (in virtual reality) to browse the library shelves from outside the library.
We have also tried to address the problem of providing clear paths in the “flat” world of the Web, where everything is potentially within a click’s reach[5].

2.      The capacity to produce documents which either use or reference other online available material.
This is how online, innovative music analyses (see figure 5) are realized by musicologists, as well as educational tools based on contemporary music pieces (see figure 6) using synchronization of score and performance, and inclusion of other material found in the online archives.


Zone de Texte:  
Figure 4: a catalog record with links to the composers' database and to multimedia contents

The online availability of large collections (of audio, of video, of text) provides a good laboratory for research in techniques of automatic indexation of multimedia contents. Such work is currently in progress in the Research and Development department of IRCAM, mainly in conjunction with the development of MPEG-7 within several European projects. As to the Multimedia Library, it takes an active role in the music information retrieval series of symposia (as member of the organizing committee, as hosting its mailing list) and to its related community of researchers and librarians.

We have strong hopes that these activities will contribute to address all the remaining issues which I mentioned above as to providing adequate specialized search tools for the library patrons.

Zone de Texte:  
Figure 5: a derived document, integrating the score synchronized with a recording of a performance, and including two spoken excerpts from the audio archives

Zone de Texte:  
Figure 6: excerpt of the musical analysis of Jupiter, by Philippe Manoury

[1] Institut de Recherche et Coordination Acoustique/Musique

[2] MUSIC-IR, for music information retrieval; BIBLIOMUS, for music librarians, etc. See the Web page for more information.

[3] Functional Requirements for Bibliographical Records, as laid out by IFLA (International Federation of Library Associations and Institutions)


[5] See our article « Le site Web de la bibliothèque considéré comme un espace », in Bulletin des bibliothèques de France 2000 t. 45 n° 3, mai 2000 (available in the Web at

Serveur © IRCAM-CGP, 1996-2003 - document mis à jour le 20/06/1997 à 11h03m40s.