Vmap. A Dublin Core Application Profile For Musical Resources

VMAP. A Dublin Core Application Profile for Musical Resources Carlos A. Iglesias1 , Mercedes Garijo2 , Daniel Molina1 , and Paloma de Juan2 2

1 Germinus XXI (Grupo Gesfor)? Depto. Ingenier´ıa de Sistemas Telem´ aticos, Universidad Polit´ecnica de Madrid??

Abstract. This paper details a Dublin Core Application Profile defined for cataloguing musical resources defined within the European eContentPlus project Variazioni.The metadata model is based on FRBR and has been formalised with DC-Text and implemented in an available web portal where users and music institutions can catalogue their musical assets in a collaborative way.

1

Introduction

According the Lynch [1], the new context for bibliographic control in the new millennium will certainly reorder priorities for investment in bibliographic control practices and will change the way that cataloguing information is used. He identifies three general approaches to identifying potentially relevant information: through bibliographic surrogates that represent an intellectual analysis and description of aspects and attributes of a work; through content-based techniques that compare queries to parts of the actual works themselves (or to computationally-derived surrogates for the works); and through social processes that exploit the opinions and actions of communities that author, read, and evaluate works, and the information seeker’s view of those communities of people involved. This paper details a Dublin Core Application Profile (DCAP) defined for cataloguing musical resources defined within the European eContentPlus project Variazioni. Variazioni project aims at enriching musical contents provided by musical institutions and end-users by combining the three processes: manual cataloguing based on well-defined metadata by musical institutions and end users; automatic cataloguing based on audio analysis, and social tagging. This paper is focused on the definition of the metadata schema for manual cataloguing The article is structured as follows. First, section 2 reviews the limitations of existing metadata standards and projects for cataloguing musical assets. Then, ?

??

This research has been co-funded by the European Community under the programme eContentPlus. The authors are solely responsible for this article and it does not represent the opinion of the European Community. The European Community is not responsible for any use that might be made of information contained within it. This work has been partially funded by the Spanish Ministry of Industry, Trade and Tourism under the Avanza Programme in the project Musiteca.

section 3 presents the model of the Variazioni Musical Dublin Core Application Profile (VMAP). Finally, section 4 draws out some conclusions and future work.

2

Limitation of metadata standards for Musical Assets

After reviewing relevant metadata standards in different domains related with music assets (Libraries, Museums, Education, Audivisual and Music), the first conclusion is that any of the reviewed standards deal with the cataloguing of music resources with enough detail for fitting user requirements in terms of search facilities and collocations. In addition, there are important limitations in traditional cataloguing systems for music resources Traditional library cataloguing records, based on AACR2R [2] cataloguing rules and MARC [3] bibliographic and authority standards have provided a solid foundation for the required descriptive metadata elements for searching and retrieving works of music and are used by music cataloguing agencies worldwide [4]. Nevertheless, they present limitations for the music domain [5,4]. First, they lack of adequate structural and administrative metadata for expressing the internal structure of the musical objects. In addition, they consider only the role of author, while other roles such as peformer or composer which are relevant for grouping results for end users are not considered. Other examples include limitations with just one title (track title, CD container title, alternative title, etc.) and with just one date (date of performance, composition, record creation, etc.), or the lack of an object oriented metadata model which provides facilities for including key entities such as composition, performer or composer. FRBR (Functional Requirements for Bibliographic Records) [6] has accomplished a shift in the cataloguing area, putting emphasis on a conceptual model which is focused on the Work rather than on the Manifestation. FRBR has been applied previously in the musical domain, and new library standards, such as RDA [7] or IAP [8] are based on FRBR. Our conclusion is that FRBR is a good starting point for defining and modelling Variazioni metadata. This conclusion could be considered in a wider scope. According to Gartner [9], “given the complexity of metadata requirements, it is perhaps not surprising that no single standard has yet emerged which addresses them all. Nonetheless, the emergence of the standards detailed in this report, all of which are based on the Functional Requirements for Bibliographical Records (FRBR) conceptual model, and the interoperability allowed by their common language, does allow for a coherent metadata landscape to be constructed on a sector-wide basis.” METS [10] and MPEG-21 [11] are two standards which attempt to provide overall frameworks within which descriptive, administrative and structural metadata. and have emerged from different communities [9]. While METS comes from the library community (the MARC standards office), MPEG-21 comes from the multimedia community. The project Variazioni counts with experts in MPEG-21, and the resulting metadata will be available in MPEG-21. Regarding the standards developed in the museum community, they deal with aspects not relevant for Variazioni (physical location or provenance of the

items) and, in addition, there is an adaptation of FRBR, so-called FRBRoo, which provides an effort in modelling CIDOC CRM [12] based on FRBR entities. The usage of controlled vocabularies will be discussed in the presentation of the Variazioni metadata model definition. Another interesting specification is IAP (Image Application Profile) [13,8], which defines an application profile based on DC, FRBR, VRA [14], CDWA [15] andMIX [16]. The basic FRBR model is modified. For sake of clarity, FRBR Work is renamed as Image, in order to distinguish the common use of work in image cataloguing (the image, physical thing, manifestation) from the abstract Work defined in FRBR. In addition, expression is omitted in this profile, since it is not considered useful for cataloguing images. In the musical domain, the most relevant projects are Variations [17,5,18,19], Music Australia [20,21] and Harmos [22]. Harmos was an eContent project which preceded Variazioni and was focused on cataloguing master classes. The educational taxonomy for cataloguing master classes [22] has been included in this model. Variations and MusicAustria have used FRBR as their basis. According to Riley [23], FRBR is the most influential model and represents a great deal of potential for search systems based on high-quality, structured metadata. Several FRBR-like systems have been implemented for the music domain, such as Variations2 [Not04] from Indiana University, the British Library Sound Archive [Not04] Opera Archives [Pom05] and MusicAustralia [MAUrl]. The application of FRBR [24] in the next version of AACR2, called RDA, has provided some findings, such as 78% of works have only one manifestation, which could question its benefits. Nevertheless, “works in fields like music, that appear in multiple versions over time, will benefit the most from systems that implement FRBR principles” [24]. Buchanan [25] has shown how users and librarians can benefit from the exploitation of FRBR relationships, as well as the inclusion of new services, such as alerts. The music community has been very active in discussing modelling issues [26]. One of the problems that has been pointed out is how to deal with Music and Lyrics in vocal music, which is the work (the music or the lyrics or both)?1 A second problem is how FRBR model fits in non Western music tradition, where there is not a previous Work. Variations is focused on a digital music library, while our project addresses a wide range of music contents, which has lead to distinguising different kinds of musical contents and its physical embodiment in productions. The Music ontology [27] is also based on FRBR, but, as well as MusicBrainz[28], they are also focused on digital music libraries, which are not the only focus of our project. Music Ontology has been considered very interesting and as future work a mapping between our model and this ontology could be defined. Some of the concepts of the Music ontology has been considered too complex for this model. For example, a Composition is an Event which produces the Musical Work. Variazioni Metadata model can contribute to this ontology refining some of their concepts.

3

Metadata modelling for musical resources using FRBR

In this section it is discussed how the FRBR conceptual model can be applied for cataloguing different types of musical contents, such as master classes videos or digitalised scores. In order to understand better the relationship with FRBR, a first identification of FRBR entities per musical content type has been carried out as shown in table 1.

Variazioni Contents Master class Score Concert Image* Studio Recording Libretto

FRBR 1st Group Entities Work Expression Manifestation Master Class Class Event Production Composition Editorial Event Printing Composition Concert event Production Image itself “Event” Production Composition Production Event Production Composition, Editorial Event Production Lyrics Table 1. Identification of FRBR entities

Item Media Media Media Media Media Media

File File File File File File

From the previous table, the following issues can be pointed out. – Expression and Work entities are not easy to identify in some cases, such as Master Classes or Conferences. This happens because the intellectual or artistic activity (Work) emerges while the activity (Expression) is being carried out. A similar issue has been previously reported for Western Music or Jazz improvisation in FRBRList [26] or MusicAustralia [20]. – According to FRBR, an Expression is the realisation of one and only one Work [6]. This can create some problems while cataloguing if the final digital file contains several Expressions (for example, a video recording with several performances or a digitalised score book with several scores, or a CD in only one track) and there is not a segmentation tool available in the system. The main Work entity in the music domain is Composition. Nevertheless, in some musical contents, such as Master Classes or Conferences, the Composition is not the intellectual / artistic activity of the Master class / Conference, but it is commonly used to exemplify a concept. They are used as subjects. – Managing image and ’event material’. The image content is problematic. For example, let us consider a concert, where there are a video recording, an audio recording and photos of the event. One natural alternative is considering that all of them are ’Manifestations’ of the same Expression (the Concert) but recorded in different media (image, video or sound). The main problem is that the photo may not be easily linked to the performance of one particular Work, but to the general event. A similar case happens for cataloguing related material such as the announcement poster of the Concert. According to [29], these augmentations (illustrations, notes, glosses, etc.) of the Expres-

sion should be considered separate Expressions of their own separate works, but this makes the cataloguing more laborious. – In digital libraries, the distinction between Manifestation and Item is not so relevant, since there is only one copy of the work (the digital media). FRBR cannot be considered as a data model, but a conceptual schema. FRBR does not even require implementing the four entities of the first FRBR Group [29]. – While FRBR follows a top-down approach for cataloguing, cataloguing follows a bottom-up approach. Users or librarians catalogue an Item, not a Work. Users should have an easy interface in order to catalogue their media files, without being aware of the FRBR model. Expertise in implementing FRBR in standard databases [20] has shown its utility for end users to find relationships between items, which were hidden before its implementation. Nevertheless, these experiences have shown that since FRBR provides several alternatives during the cataloguing process, this can make the process difficult to understand. Some examples of these difficulties are to decide whether music and lyrics should be catalogued as different items, the definition of relationships between expressions (i.e. an interpretation (e1) based on a libretto (e2) of a work (o1)), as the cataloguing of expressions based on improvisation, such as jazz music and folk traditions. – Cataloguing is commonly carried out in an iterative way in musical institutions. Depending on the available resources, a media file can be catalogued with very few metadata and catalogued more exhaustively when there are more resources. This include the identification of the Composition (Work), which can be delayed.

Fig. 1. VMAP and FRBR entities

Based on the limitations identified above, an adaptation of FRBR for musical resources is here proposed. This model, Variazioni Musical Application Profile (VMAP) is based on the existing FRBR entities, although these entities has been renamed and redefined as shown in figure 1, in order to overcome the identified limitations. In particular: – Work has been limited to Composition. A Composition is an original piece of music. – Expression has been redefined as Musical Content. A Musical Content (Musical Content Type) is a classification scheme of digital items which defines the nature and descriptive metadata of the digital item. Some of the musical content types identified are Master Class, Conference, Libretto, Musical Score, etc. – Manifestation has been renamed as Production. A Production maintains all the metadata related to the physical edition of a Musical Content, as well as the structural metadata when the manifestation is composed of more than one Media Fragment. The structural metadata can include the order of different Media Fragments or the starting and end points of one media file with different fragments (pages, seconds, frames, etc.). Two subtypes of Production have been identifed, document and audiovisual, for defining specific metadata. – Item has been renamed as Media Fragment. A Media Fragment is a media file or a fragment of it, and maintains all the relevant metadata of the media file, including its title and licence.

W1. J. S. Bach’s Six suites for unaccompanied cello E1. Transcription for classic guitar by Stanley Yates M1. Publication of the guitar transcription by Mel Bay Publisher in 1988 I1. Exemplar of the book in library 1. I2. Separata of the guitar edition in library 1. E2. Performances by Janos Starker recorded in 1963 and 1965 M1. Recordings released on 33 1/3 rpm sound discs in 1965 by Mercury M2. Recordings re-released on CD in 1991 by Mercury Table 2. FRBR modelling. Legend: W: Work, E: Expression, M: Manifestation, I: Item

In order to clarify these elements, an example of how the same items are catalogued according to standard FRBR and Variazioni Music Application Profile is shown in tables 2 and 3 respectively. From this example, the main differences of the model can be outlined. First of all, according to FRBR, and Expression has one and only one Work, and this has supposed the shift in focus from the resource (Manifestation) in the traditional cataloguing world to the Work in FRBR. Our proposal consists of modifying

MC1. Score. Transcription for classic guitar by Santley Yates C1: J. S. Bach’s Six suites for unaccompanied cello P1: Book edition MF1: Media file of the book (page range if book includes more compositions) P2 Separata of the guitar edition MF2: Media file of the separata MC2. Studio Recording. Performances by Janos Starker recorded in 1963 and 1965 C1: J. S. Bach’s Six suites for unaccompanied cello P3: Recordings released on 33 1/3 rpm sound discs in 1965 by Mercury. MF3: Suite 1 media file (and details of the fragment, full or time range) C2: J. S. Bach Suite 1 for unaccompanied cello [is-part-of C1] MF4: Suite 2 media file (and details of the fragment, full or time range) C3: J. S. Bach Suite 2 for unaccompanied cello [is-part-of C1] ... MF8: Suite 6 media file (and details of the fragment, full or time range) C7: J. S. Bach Suite 1 for unaccompanied cello [is-part-of C1] P4: Recordings re-released on CD in 1991 by Mercury MF9: Single media file of the suites Table 3. VMAP modelling. Legend: C: Composition, MC:Music Content, P: Production, MF: Music Fragment

the cardinality of the relationship hasWork between Work and Expression, from 1-1 in FRBR to M-M (many-to-many). This allows solving some of the previous issues pointed out: (a) , since Compositions (Works) are not mandatory for a Musical Content (Expression); and (b), since one Musical Content (Expression) can have more than one associated Compositions (Works). Another interesting change is the usage of the relationship hasSubject, in particular for linking any element of the model with Composition. FRBR only considers this relationship for Works. In our case, for example, for Master classes, several Compositions could be the subject (or example) of a master class. In the example previously presented, a Composition can be assigned as subject of a Media Fragment, avoiding the need for creating a new Expression. This is depicted in figure 1, which points out two different kind of semantic relationships between Composition and Musical Content: isRealizedAs and hasSubject. In terms of searchability, we have not found the need to distinguish between both in the implementation of the model. Furthermore, it is possible to define the subject of a media fragment, allowing a direct asignation The relationships between the entities of the VMAP model are shown in figure 2. It is important to point out the cardinality between Composition and Musical Content. A Musical Content can have more than one Composition (as a realisation or as a subject). In addition, Media Fragments can have one or more Compositions as a subject or realisation. In order to simplify the cataloguing process, it is not distinguished between the relationships isPerformedAs and hasAsSubject, shown previously in figure 1.

Fig. 2. VMAP model

It is important to point out, that VMAP considers MusicalContent as the central entity. The entity Composition can be identified when there are resources to do this catalogation, but it is not required, in contrast with FRBR, where the entity Work should be identified in first place. According to the requirements of the content providers in the project, the following musical contents have been identified: Score: Musical score, it can be an scanned handwritten, autograph or printed score or a computerised score document. Musical document: Document related with the musical domain, such as image, libretto, literature on music, reference works, correspondence between musicians, methods, press releases or concert programs. Class: Audiovisual content about some pedagogical activity, such as master class or regular class. Conference: Audiovisual content recording of a conference about music. Musical performance Audiovisual content with the recording of a musical performance, such as concert, competition or musical event. For each musical content type, specific metadata have been identifed as shown in figure 4 and is formalised in [30]. This model has been formalised as a Dublin Core Application Profile for describing digital musical contents hold by musical institutions or end users. The concept of Application Profile has emerged within the Dublin Core Metadata Initiative. Dublin Core Application Profiles (DCAP) can be defined [31] as schemas consisting of data elements drawn from one or more name spaces, combined together by implementers, and optimised for a particular local application

Fig. 3. Musical Content Entity

Dublin Core Application Guidelines [32] and the Singapore guidelines [33] have been followed to document attributes drawn from other name spaces and the declaration of new elements under the name space VMAP (Variazioni Musical Application Profile) The resulting DCAP is available at [30] and has been formalised with DCText [34], where all the entities and properties have been formalised, as well as 26 classification schemes. VMAP reuses several controlled vocabularies, such as Getty Thesaurus for Geographical names [35], the thesaurus of Musical Instruments [36] as well as metadata schemas such as FOAF [37] In addition, a mapping to Simple Dublin Core has been defined in order to provide OAI-PMH interoperability. The model has been implemented following the aspect oriented content model of Alfresco [38] and is publicly available at the Variazioni Content Enrichment Model (CEP) [39].

4

Conclusions and Future work

This article has reviewed the practices in cataloguing musical contents and has proposed a Dublin Core Application Profile for Musical Digital Contents. The usage of FRBR as a basis for the musical metadata mode has provided a suitable model, although the need to redefine the entities, could question the suitability of this model and its interoperability, since several Dublin Core Application Profile have followed the same approach in redefining core FRBR entities,

Fig. 4. Subtypes of Musical Content Entity

as discussed in [40]. This could suggest the need of defining a more extensible model based on FRBR, which would improve the interoperability. Although VMAP has been defined for the musical domain, it could be applicable in other arts, which is left to future work. Defining a metadata schema which provides quality metadata is not an easy task. It enables a trade-off between different quality properties, such as consistency, completeness, accuracy, shareability, economic feasibility and usability. One could think that the most complete and more metadata a schema has, the more quality it has. Nevertheless, one of the common pitfalls pointed out in [41] for providing shareable metadata is too much information. Economic and human factors are also important in the definition of metadata, in order to provide understandable and effective metadata. Furthermore, natural resistance to change is a challenge for adopting a new metadata model. In order to define quality metadata, the process defined in [42] has been followed for defining Variazioni Metadata Model, identifying external and internal requirements. The implementation of the project has provided feedback about the decisions taken in this definition. Currently partners of Variazioni are working in the dissemination of the VMAP model as well as the portal. Feedback from users will guide its evolution. In addition, the Spanish project Musiteca has started a living lab for classic musical contents where this DCAP will be validated with end users and classical music content providers.

References 1. Lynch, C.: The new context for bibliographic control in the new millennium. Library of Congress January, 2001. Available at: http://www.loc.gov/catdir/bibcontrol/lynch paper.html (2001) 2. Hunter, E.J.: AACR 2 : an introduction to the second edition of Anglo-American cataloguing rules. London : C. Bingley ; Hamden, Conn. : Linnet Books (1979) 3. of Congress, L.: Marc (machine readable catalogue) web site. Available at http://www.loc.gov/marc/ . 4. Hemmassi, H.: Why not MARC? In: Proceedings of the 3rd International Conference on Music Information Retrieval, ISMIR. (2002) 242–248 5. Minibbayeva, N., Dunn, J.W.: A digital library data model for music. In: Proceedings of the Second ACM/IEEE Joint Conference on Digital Libraries. (2002) 154–155 6. IFLA: FRBR functional requirements for bibliographic records. Technical report, International Federation of Library Associations and Institutions, IFLA (1998) Available at http://www.ifla.org/VII/s13/frbr/frbr.pdf. 7. for Development of RDA, J.S.C.: RDA: Resource description and access. Available at ”http://www.collectionscanada.gc.ca/jsc/rda.html#drafts” 8. JISC: Images application profile Available at http://www.ukoln.ac.uk/repositories/digirep/index/Images Application Profile. 9. Gartner, R.: Metadata for digital libraries: state of the art and future directions. JISC Technology and Standards Watch (2008) 10. Digital Library Federation: Metadata encoding and transmission standard: Primer and reference manual. (September 2007) 11. Wan, X.: Mpeg-21 rights expression language: enabling interoperable digital rights management. Multimedia IEEE 11(4) (October-December 2004) 84–87 12. of Museums, I.C.: Cidoc web site. Available at http://cidoc.ics.forth.gr/ 13. Eadie, M.: Towards an application profile for images. Ariadne Electronic Magazine (55) (April 2008) Available at http://www.ariadne.ac.uk/issue55/eadie/. 14. of Image Media Professionals, I.A. 15. Trust, J.P.G.: Categories for description of works of art. Available at http://www.getty.edu/research/conducting research/standards/cdwa/index.html (2006) 16. NISO: MIX. NISO Metadata Standard for Images in XML Schema. technical metadata for still images standard. Available at http://www.loc.gov/standards/mix/ 17. Dunn, J.W., Mayer, C.A.: Variations: A digital music library system at indiana university. In: Proceedings of the 4th ACM Conference on Digital Libraries, Berkeley, CA (1999) 12–19 18. Jenn Riley, Caitlin Hunter, C.C., Berry, A.: Definition of a frbr-based metadata model for the indiana university variations3 project. Technical report, University of Indiana (2007) 19. Jenn Riley, Casey Mullin, C.C., Berry, A.: Definition of a frbr-based metadata model for the indiana university variations3 project. phase 2: Frbr group 2 and 3 entities and frad. Technical report, University of Indiana (2008) 20. Ayres, M.L.: Case studies in implementing functional requirements for bibliographic records [frbr]: Auslist and musicaustralia. Australian Library Journal 54(1) (2004) Available at http://www.alia.org.au/publishing/alj/54.1/full.text/ayres.html.

21. MusicAustralia: Musicaustralia web site available at http://www.musicaustralia.org 22. Carlos A. Iglesias, Marta S´ anchez, M.J.G.A.G., G´ omez, E.: A multilingual a multilingual web-based educational system for professional musicians. In: Current Developments in Technology-Assisted Education, Proceedings of Fourth International Conference on Multimedia and Information and Communication Technologies in Education, Sevilla, Spain (2006) 23. Riley, J., Mayer, C.A.: Ask a librarian: The role of librarians in the music information retrieval community. In: Proceedings of the International Conference on Music Information Retrieval, ISMIR06. (2006) 24. Riley, J.: The essence of information technology for library decision-makers. In: FRBR 25. Buchanan, G.: Frbr: Enriching and integrating digital libraries. In: JCDL ’06. Proceedings of the 6th ACM/IEEE-CS Joint Conference on Digital Libraries. (2006) 260–269 26. le Boeuf, P.: Frbr mailing list, list address [email protected] 27. Giasson, F., Raimond, Y.: Music ontology specification. Available at http://musicontology.com/ (2007) 28. Friedich, M., Kaye, R.: Musicbrainz metadata initiative 2.1. XML metadata format 29. IFLA Working Group on the Expression Entity : FRBR Chapter 3: Entities, Proposed changes to the FRBR Text 30. Iglesias, C.A., Molina, D.: Variazioni musical application profile. Available at http://www.variazioniproject.org/vmap (2009) 31. achel Heery, Patel, M.: Application profiles: mixing and matching metadata schemas. Ariadne Electronic Journal (25) (September 2000) 32. Thomas Baker, Makx Dekkers, T.F., Heery, R.: Dublin core application profile guidelines. Available at http://dublincore.org/usage/documents/profileguidelines/ (2005) 33. Nilsson, M., Baker, T., Johnston, P.: The singapore framework for dublin core application profiles. Available at http://dublincore.org/documents/2008/01/14/singapore-framework/ 34. metadata using the DC-Text format, E.D.C.: Pete johnston. Available at http://dublincore.org/documents/dc-text/ (2007) 35. Trust, J.P.G.: Getty thesaurus for geographical name web site. Available at http://www.getty.edu/research/conducting research/vocabularies/tgn 36. Jenkins, K.: Thesaurus of musical instruments. Available at http://alteriseculo.com/instruments/. (2003) 37. Bricklye, D., Miller, L.: Foaf vocabulary specification 0.91, namespace document 2, openid edition. Available at http://xmlns.com/foaf/spec/ (November 2007) 38. Alfresco: Alfresco web page. Available at http://www.alfresco.com 39. project, V.: Variazioni content enrichment portal. Available at http://cep.variazioniproject.org 40. Chaudhri, T.: Assessing frbr in dublin core application profiles. Ariadne Electronic Magazie (58) (jan 2009) Available at http://www.ariadne.ac.uk/issue58/chaudhri/. 41. Sarah L. Shreeves, J.R., Milewicz, L.: Moving towards shareable metadata. First Monday 11(8) (August 2006) Available at http://www.firstmonday.org/issues/issue11 8/shreeves/index.html#s2. 42. Marieke Guy, A.P., Day, M.: Improving the quality of metadata in eprint archives. Ariadne (38) (January 2004) Available at http://www.ariadne.ac.uk/issue38/guy/.