Abstract

This article presents the Hourglass Model, a theoretical framework allowing to redesign the strategies for the creation and use of mainly descriptive metadata in audiovisual archives. The model outlines four main categories of metadata creation (manual annotation by archivists, metadata coming from production, user-generated metadata, and automatically extracted metadata) and three main categories of use (preservation and collection management, search and retrieval, and enhancement and contextualization). Archives can plan an effective and efficient metadata strategy by connecting one or more creation methods with one or more purposes. Next, recent developments in the creation and use of descriptive metadata are discussed. The fact that these recent developments can be fit in easily show that the model is future-proof. Furthermore, it is suggested that the implementation of the Hourglass Model has consequences for the archive not only on the technical level, but also on the institutional and organizational ones. It stresses the importance of archival openness as an institutional leitmotiv, a new role for the archivists from the organizational perspective, and modular and exchangeable services as a key characteristic of the archive’s IT infrastructure.

Keywords

archives, audiovisual media, documentation, cataloging, metadata, strategy.

Introduction

The quantity of audiovisual media (audio, video, and film) to be processed by audiovisual archives is globally in a constant increase. The growth of born digital items is accelerating and additionally digitization projects are going on, often aiming to digitize the entire collection of an archive (just a few recent examples: Monette, 2015; Messmer and Houpert, 2015; or Casey and Merten, 2015). Of the objects to be digitized large parts are not or only very superficially described or cataloged.

Metadata (descriptive, administrative, technical) are indispensable for archives in general for several key tasks such as storage management, preservation and making the collection searchable. In digital audiovisual archives in particular, this indispensability is by no means lower. Moreover, the user groups asking for access to these collections are becoming not only larger but also more diverse. New (actual or potential) user groups of the audiovisual archive additionally have their own needs in terms of access. Not only when it comes to the interface features, but also in metadata: types, form, vocabulary and level of detail in the descriptions (de Jong, 2010). Finally, television and radio broadcasters managing huge quantities of media are shifting their focus from the delivery challenge to the discovery challenge.¹ Also this factor stresses the essential role of descriptive metadata.

The history of documentation of audiovisual archives—or should we say audiovisual content in general?—was dominated until the 2000s by manual description by archivists. This method is known as particularly time-consuming, taking five times the duration of the media itself easily. But since then there has been a rapid evolution underway in the area of alternatives to the manual creation of metadata. In addition, there is an equally rapid evolution in the field of the use of these metadata. Previously metadata were quite commonly divided into administrative, technical and descriptive metadata, but since the second half of the 2000s terms as automatically extracted metadata, user generated metadata, preservation metadata, linked metadata, and so on have become ever more common, albeit mainly in theoretical literature and proofs of concept of all kinds. Despite this rapid conceptual change a big delay can be observed in the adoption in daily practice of the processes behind these terms.² On technological, organizational as well as institutional levels audiovisual archives seem to have difficulties with their implementation.³

The Hourglass Model

The Hourglass Model has the ambition to serve as a theoretical framework for developing strategies for the creation and use of (descriptive) metadata. Originally developed in 2013 for the archives department of the Danish public broadcaster DR, it describes the place that metadata hold in contemporary audiovisual archiving, starting from the question how the different types of metadata relate to each other.

Preservation metadata for example, are sometimes automatically extracted. Still, no one will consider them synonymous terms. When analyzing the terms, it appears that some terms describe the way metadata are created, while others describe the purpose that they are used for. This leads us to situating them on different levels (fig. 1): a creation level, grouping the ways in which metadata are created or collected, and a purpose level, describing the goals that the metadata serve. These two levels are connected by archival processes, on which we will come back later on in this article.

Creating and collecting metadata and associating them with their corresponding essence today take very different forms. If we take a closer look however, it shows that the most significant differences between the methods of creation are in the moment of their creation (i.e., the point in the audiovisual value chain), and who or what is their creator or collector (fig. 2).

In fact, the creation methods can be divided into four different groups: first, there is the manual annotation by archivists, ideally right before giving the item its final place in the archive (but often much later), but at least after the production of the media object. Secondly, there are production metadata, a group holding a wide variety of metadata, which can be created both human and automatically. In any case they are created before entering the archives, during the production phase⁴, in many cases with other purposes than strictly archival ones.

User-generated metadata are a third group of metadata creation methods. These are metadata created by the user, at any given moment, consciously or unconsciously, and through any technology. Finally there are automatically extracted metadata: they are the outcome of a broad range of technologies allowing the content of sound, moving pictures or still images to be described automatically without human intervention other than applying a specific algorithm to the media essence. Broadly speaking, these technologies can again be divided into classification, segmentation, recognition, identification, and transcription. The result can consist out of a text (such as a transcript of what is said in the media) but also for example out of a series of time codes indicating meaningful changes in the media, for example the transition of human speech to music or a new person coming into the picture. These transitions are often displayed in the user interface of the access platform⁵ in the form of markers on a timeline.

At the purpose level we find the final objectives for which these metadata are used. We stress the word “final” here, because the human or automatic retrieval of the essence is rarely a goal in itself, but nearly always the starting point of a further purpose. Analyzing these purposes we can in fact identify three of them, the first being preservation and collection management. Evidently the term preservation metadata rises here: data about provenance, quality control (whether or not after one or more migration processes in any form) and all other kinds of actions of file management that the media object undergoes throughout its lifecycle.

Second, there is the most classical focus of archival disclosure: search and retrieval for reuse in whatever context. And finally there is enhancement and contextualization: linking the content from one collection with other content (where also the concept of linked metadata can be situated).⁶ The most obvious use of the Hourglass Model involves a way of creating metadata being associated with a purpose that one wants to achieve. In fact, any of the four ways of creating metadata can be combined with any of the three purposes. The model thus results in twelve interesting combinations, some of which have been known for centuries (using full manual annotation for managing the collection for example) and others were born only recently. Based on the purpose, some form of metadata creation is chosen, taking into account the characteristics of the essence and the strengths and weaknesses of the creation method. It is therefore important that the archivist is well aware of these aspects. For example:

A radio broadcaster’s archive wants to connect a series of undescribed interviews with politicians with pictures of the same politicians from other collections, via an online platform. To save time, the annotation of the interviews is not done manually, but by making use of the introduction texts written by the journalists while preparing their broadcasts. A disadvantage of these texts that the archivist should be aware of is that radio journalists often write names phonetically, instead of using the correct spelling. This will evidently obstruct the automatic connection of the collections.

Nothing however should stop the archivist from using different annotation methods for different parts of the collection, or to make sure that one source of metadata can serve multiple purposes. Furthermore, it could be interesting to cope with the weaknesses of one method, via the strengths of another one. A good example of the combination of metadata creation of all four categories, is the BBC World Service Archive Prototype (Raimond, 2013).

Extending the Model

In recent years, both on the creation level and on the purpose level a clear innovation and further detailing of existing methods can be noticed. On the creation level these have corrected some weaknesses as they were known before. On the purpose level they furthermore allow for new possibilities within the three groups of reuse listed above.

With regard to production metadata, two evolutions are clearly visible. On the one hand there are some recent technological innovations. Production platforms (sometimes referred to as a production asset management system [PAM]) keeping track of the essence in various phases of the production process ever more acknowledge the importance of descriptive metadata already at that stage of the media value chain. Capturing metadata at that point in a structured and uniform way is beneficial not only to the efficiency of the production, but also later on during the archiving processes (Verwaest, 2014). On the other hand, the digitization of paper archive collections and the (semi-) automatic extraction of their content, serving as descriptive metadata is ever more common. BBC’s Genome Project is a nice example of this trend (O’Dwyer, 2012). This shows that the use of production metadata is not limited anymore to recently produced content and that the full collaboration of the producers is not always required. If they can be harvested easily, the reuse of production metadata is the low-hanging fruit of metadata creation for the archive, offering often quite detailed metadata in an inexpensive and quick way.

Regarding user generated metadata an interesting distinction can recently be made between the conscious and unconscious creation of metadata. Consciously creating user generated metadata is usually associated with crowd-sourcing, though it does not necessarily have to be located there. It can be done in a high-tech way (Noordegraaf, 2012) as well as in a way that requires little technological investment. The Irish public broadcaster RTÉ managed for example to cancel out a number of question marks from its catalog via Twitter calls to the general public. The fact that this project reinforced the archive’s connection with the general audience was more than just a positive side effect (Wylie, 2013).

As metadata created unconsciously by the user we consider data about the essence collected mostly automatically during the use or reuse (i.e., consumption) of the media, without the user taking specific action for this to happen, often even without being aware that this is happening. Since a few years already the term consumer-generated metadata is being used for this (Rule, 2008). In a certain sense, television audience ratings for example can be considered consumer generated metadata. But also other data on the perception of the media by the public can be considered metadata of interest to the archive. Monitoring and analysis of the viewer’s emotions, based on second screen social media activity for example, is an ever-more common part of audience research, done or commissioned by broadcasters. In more general terms, we can state that particularly in broadcasting the archive and the audience research department are increasingly partners. Whether it is about use statistics or specific characteristics of the content, they ever more become customers and suppliers of each other’s data. Since audiovisual archives are increasingly becoming platforms for media access, a thorough monitoring of customer behavior is also a matter of self-interest for the archive. Not only these traces of media consumption enable them to match supply directly with customer demand, it also makes their collections better described in the longer run.

It is even possible to create user-generated metadata via crowdsourcing, without the collaborators being aware that what they consciously create, actually serves as metadata. Collaborative subtitling platforms such as Amara are originally not intended to generate descriptive metadata, but for captions, translations and subtitling (Jansen, 2015). However, it is evident that these can also serve as descriptive metadata, just as the subtitling made by professionals. In summary user generated metadata in all their new forms are a very promising form of descriptive metadata. Some older objections, such as the doubts about reliability, the risk of vandalism and issues of labor intensity are virtually gone.

In the field of automatically extracted metadata two particularly important and partly linked evolutions can recently be observed. First and foremost this way of creating metadata (i.e., metadata generated via automatic speech recognition) has made its long-awaited entry in the daily practice of audiovisual archives. At the Swiss Italian-speaking public broadcaster RSI for example the technology was commissioned in 2012 as part of a new media asset management (MAM) system (Aziz et al., 2013). The second evolution, providing automatical extraction of metadata via a software-as-a-service (SaaS) model and an application programming interface (API)⁷, has also led to a series of other implementations, usually on a project basis, but still in daily practice: the Studs Terkel Radio Archive calls on the API of Audiosearch (Schein, 2015) and the French-speaking Swiss public broadcaster RTS uses the Vocapia service for the annotation of its digitized audio tape collection (Lorenz, Seigneur and Wenger, 2015). On the SaaS market it is very easy to find other APIs offering various other technologies such as recognition or even identification of persons or objects.

Innovations at the purpose level often show a rather slow pace when it comes to MAM systems⁸, but platforms targeting the end user tend to evolve quicker, as they’re often built ad hoc, for particular projects and/or target groups. Three particularly interesting applications draw our attention.

Regarding preservation and collection management, harvesting the metadata generated by one or more devices in the digitization chain allow for a constant monitoring by the customers of outsourced digitization projects, as was demonstrated by Lorenz et al. (2015). Regarding search and retrieval the increasing quantities of essence and metadata pose a particular challenge. Search queries tend to have more results, but not per se more accurate ones. Graphic user interfaces (GUI) should therefore avoid overwhelming the user, by making use of features like improved relevancy ranking or refining results via facets. Also the idea of a slide switch meets this requirement: if a user chooses for less, but more accurate results, only the results of very reliable metadata sources are shown. To get more, but potentially also less accurate results, the search algorithm makes use of a gradually increasing number of metadata sources, also the less reliable ones. The results from the second option may also meet the need for a more serendipitous way of searching through media collections (Sauer, 2016).

Last but not least, descriptive metadata are used for innovative ways of enhancement and contextualization. Semantic linking is on the radar of audiovisual archives since a little less than ten years. It goes beyond the scope of this article to explain the functioning and the full possibilities of semantically linking audiovisual archive collections, but we will suffice by pointing to one of the more spectacular proofs of concept set up in recent years: RTBF’s GEMS project, as explained by Jacques-Jourion (2013). For this project, the available metadata from more traditional archival databases were complemented by keywords spotted in texts generated by a speech-to-text engine. These were linked to each other and to external knowledge gathered in Linked Open Data sources such as DBpedia via an Open Service Bus (OSB). The result is a web interface allowing serendipitous browsing through the media archive by journalists and revealing even unexpected links between keywords and media fragments.

Implementing the Model

To implement the Hourglass Model in an audiovisual archive is not only a matter of installing new technology, just as a MAM system is more than just a new database. The Hourglass Model can fully come into its own only if new technologies to create and use—usually designed as services around the archival database—are mirrored into the archive’s strategy and its organizational structure.

Metadata strategies in line with the Hourglass Model assume the acceptance and even the full embrace of openness as an institutional leitmotiv, leaving behind the old image of archives as fortresses of protected knowledge and archivists as omniscient gatekeepers. Metadata creation methods as well as their purposes are and will become even more linked to the outside world in the future. In this regard the Hourglass Model fully connects with what Oomen (2014) calls “open, smart and connected audiovisual archives.” The main emanation of this openness at the creation level is clearly in accepting the user as a legitimate creator of descriptive metadata⁹, while at the purpose level the output of the archive should connect also to the outside world, for example by considering the use of the archival content as the true value of it, as professed by Golodnoff and Lerkenfeld Smith (2011).

On the organizational level, implementing the Hourglass Model mainly translates into new roles for the archivist. Standing in the middle of it, at the process level, he or she should be aware of the new technological possibilities and the new institutional context of the archive, but also of the new needs of old and new target groups. Although it is likely that traditional manual annotation will still survive for a while, recognizing and connecting creation methods to purposes, while managing these processes and checking metadata quality will become ever more important. Knowing and respecting the strengths and weaknesses of each creation method and the characteristics of the collection to be annotated will help the archivist to make the most effective and powerful combinations.

When it comes to the archive as an IT architecture, the Hourglass Model is particularly suited in a flexible environment, where systems are modular and exchangeable. It assumes all but monolithic, one-purpose systems. The speed of technological evolutions together with ever more external connections calls for gradual adaptations (instead of the usual “big bang” when for example a new MAM system is taken into use) but also for solid standards¹⁰. The tendency of working with well documented micro-services clearly meets this requirement.

Conclusion

In this article we have presented the Hourglass Model, we have elaborated further on recent evolutions in metadata creation and use, and we have suggested that the implementation of the Hourglass Model has consequences for the archive on institutional, organizational and technical levels.

The Hourglass Model is not a concrete tool, but a theoretical framework for everyone who’s confronted with the redesign of the strategies for metadata creation. It is a framework for helping audiovisual archives to allow their metadata creation strategy to evolve. It is future-proof, because it allows new technologies to be fitted in easily. Evidently the metadata sources are related with four kinds of metadata creators: the maker of the essence, its user, the archivist and their devices. Next to that, also the moment of creation is an interesting, distinguishing factor.

The Hourglass Model also encourages to think out of the box when it comes to create metadata. In fact archives should consider all data surrounding the media value chain (from production to consumption, conservation and reuse) as potential metadata, in an open minded way. This means: how can these data be collected, associated with the essence and used for the strategy of the archives, the prospering of the collection and the needs of its users.

For those determining the strategies of audiovisual archives, the challenge is to redesign the existing ones, often mainly based on old-school manual annotation. In this context it is important to know these new methods, including their strengths and weaknesses. Next to that, business cases should be elaborated, taking into account the content wise characteristics of the collection (e.g. the rights situation) and the environments from which the archive acquires the content (e.g. the production methods of the broadcasting industry).

Parallels between the Hourglass Model and the Open Archival Information System (OAIS) are clearly visible. Both serve as a theoretical framework with three major components, focusing on input, output and processes in between (de Jong, Delaney and Steinmeier, 2013). This suggests that audiovisual archives who work OAIS compliant will also find it easier to think along with the Hourglass Model’s logic. This way the Hourglass Model can be considered as a way to apply OAIS but limited to the descriptive metadata aspects of it.

If all you have is a hammer, everything looks like a nail. The archivist used to have only the full manual annotation method available. Soon however, this multipurpose goody will be exchanged for a whole toolbox. Then the archivist has to become acquainted with the strengths and weaknesses of each tool and to make informed decisions which one to use. Since the archivist will manually intervene less, it will also be more important to monitor the quality of the processes well. Ultimately a good overview will be required, because the ever-faster growing collection not only has to be made searchable, enhanced and contextualized but also has to be managed and preserved.

BRECHT DECLERCQ is the Digitization and Acquisition Manager at VIAA, the national audiovisual archive of Flanders, Belgium, since 2013. He is responsible for the overall digitization strategy of the Flemish audiovisual heritage. Previously, he worked for the Belgian public broadcaster VRT for almost 10 years as a radio archivist and as a project lead in several digitizations, media asset management, and access projects. He is the current Chair of the FIAT/IFTA Preservation and Migration Commission and Member of the FIAT/IFTA Executive Council. He writes, presents, reviews, and advises to several European broadcasters and audiovisual archives.

Notes

1.  In the world of linear analog broadcasting television and radio stations struggled with delivering their content to their audiences at the right time. In a world of internet based, nonlinear media consumption, their challenge is to make sure their content is discovered amongst the enormous supply of online media (Solheim and Børdalen, 2015).

2.  For example and according to our research it took no less than 18 years between the first research of speech-to-text as an interesting technology for describing audiovisual archive items and the actual first implementation in the daily practice of a Western European broadcasting archive. See Christel, Stevens and Watclar (1994), as compared to Aziz, Vassallo and Veri (2013).

3.  Gazendam (2015) reports that cataloguers at the Netherlands Institute for Sound and Vision manage to annotate roughly one-third of all audio and video material coming in, with the annotations reaching their final stage roughly two months after the first broadcast date.

4.  For example: journalist’s texts, interview transcriptions, subtitles, location data created by the built-in GPS-data in a recording device, etc.

5.  To use a generic term; in many cases this will be a media asset management system (MAM).

6.  This link could be considered not only a form of metadata use, but also of metadata creation: by linking one document to another, the first one forms a kind of descriptive metadatum for the other.

7.  Of course, this supposes that the architecture of the archive’s IT system is at least partially service oriented (SOA).

8.  Recent research by Brodie Kusa, Declercq and Stanz (2015) with almost 50 audiovisual archives showed that more than one out of three archivists called the graphic user interface (GUI) of their current MAM system “an unsolved or dissatisfying aspect,” while almost two out of three held high hopes for improvement in this regard for their next MAM system.

9.  An interesting example can be found in the Origo radio asset management system of the Norwegian public broadcaster NRK (Engels and Wettmark, 2015). The borders between users and producers of metadata are becoming very vague in this system, as journalists are allowed and even encouraged to alter metadata in the Origo interface.

10.  We refer to the work of the FIMS task force (Framework for Interoperable Media Services), a joint initiative of EBU and AMWA (Dimino, Evain, Footen and Gilmer, 2013).

References

Aziz, S.H., Vassallo, L., and Veri, F. (2013). Speech to text and semantic analysis. Our working experience. In: FIAT/IFTA Media Management Commission (Eds.), Metadata as the cornerstone of digital archiving. Changing sceneries, changing roles, part VI. Selected papers from the MMC Seminar, Hilversum, 16–17th May 2013 (pp. 126–133). Stockholm – Vienna – Dubai: FIAT/IFTA.

Bailer, W., Höller, F., and Messina, A. e.a. (2005). State of the art of content analysis tools for video, audio and speech. Rome – Paris: PrestoSpace.

Brodie Kusa, E., Declercq, B., and Stanz, G. (2015). The FIAT/IFTA MAM Survey. An interim assessment of one decade of file-based audiovisual archiving. In: FIAT/IFTA Media Management Commission (Eds.), Setting the standard in 2nd generation MAMs and metadata. Changing sceneries, changing roles, part VII. Selected papers from the MMC Seminar, Glasgow, 22–23 May 2015 (pp. 135–167). Stockholm – Vienna – Hilversum: FIAT/IFTA.

Casey, M., and Merten, M. (2015, September 30). Media Preservation at Scale: The Indiana University Media Digitization and Preservation Initiative. Presentation, 46th Annual IASA Conference. Paris.

Christel, M., Stevens, S., and Wactlar, H. (1994). Informedia Digital Video Library. Proceedings of the Second ACM International Conference on Multimedia, Video Program. New York, NY: ACM.

De Jong, A. (2010). De nieuwe toegang tot audiovisuele content. Ontwikkelingen in de netwerkcultuur. Hilversum: Nederlands Instituut voor Beeld en Geluid.

De Jong, A., Delaney, B., and Steinmeier, D. (2013). AV Preservation workflows: requirements for OAIS compliant workflow modelling. In: FIAT/IFTA Media Management Commission (Eds.), Metadata as the cornerstone of digital archiving. Changing sceneries, changing roles, part VI. Selected papers from the MMC Seminar, Hilversum, 16–17th May 2013 (pp. 62–84). Stockholm – Vienna – Dubai: FIAT/IFTA.

Dimino, G., Evain, J.-P., Footen, J., and Gilmer, B. (2013). FIMS, a Framework for Interoperable Media Services. SMPTE Motion Imaging Journal. 122(6), 93.

Engels, R., and Wettmark, M. (2015, May 22). Major changes at NRK or how to overhaul your metadata infrastructure using semantic web. Presentation, FIAT/IFTA Media Management Seminar ‘Changing Sceneries, changing roles. Part VII’. Glasgow.

Gazendam, L. (2015). Cataloguer support in cultural heritage (Unpublished doctoral dissertation). Vrije Universiteit, Amsterdam.

Golodnoff, T., and Lerkenfeld Smith, M. (2011). Exploring cultural heritage and value creation. A case study of DR’s Cultural Heritage Project (Unpublished master thesis). IT University, Copenhagen.

Lopez, O., Moens, M.F., and Matton, M. (2013). State of the art on semantic retrieval of AV content beyond text resources. Geneva: EBU.

Lorenz, T., Seigneur, J.M., and Wenger, O. (2015, October 1). Use case: large scale outsourced digitization project with automatic supervision. Presentation, 46th Annual IASA Conference. Paris.

Jacques-Jourion, X. (2013). GEMS. Semantics for audiovisual dummies. In: FIAT/IFTA Management Commission (Eds.), Metadata as the cornerstone of digital archiving. Changing sceneries, changing roles, part VI. Selected papers from the MMC Seminar, Hilversum, 16–17th May 2013 (pp. 32–41). Stockholm – Vienna – Dubai: FIAT/IFTA.

Jansen, D. (2015, December 4). Community driven video accessibility. Presentation, EUScreenXL Conference. Warsaw.

Messmer, R., and Houpert, J. (2015, October 10). Quality control in ORF Broadcast and Archive Migration Workflows. Presentation, FIAT/IFTA World Conference. Vienna.

Monette, P. (2015, October 10). Radio Canada: Digitization of the audio, video and film collections. Presentation, FIAT/IFTA World Conference. Vienna.

Noordegraaf, J. (2012). Crowdsourcing Television’s Past. The State of Knowledge in Digital Archives. Tijdschrift voor Mediageschiedenis. 14(2), 108–120.

O’Dwyer, A. (2012). Digitising Context. The Case of the Radio Times. VIEW. Journal of European Television History and Culture. 1 [1], Utrecht – Maastricht – London. Igitur Publishing.

Oomen, J. (2014, October 29). Towards more open, connected and smart audiovisual archives. Presentation, Simpósio Internacional de Preservación Audiovisual y Digital. Archivos Contemporaneos. Mexico City.

Raimond, Y. (2013). The BBC World Service. An alternative approach to publishing large archives. In: FIAT/IFTA Media Management Commission (Eds.), Metadata as the cornerstone of digital archiving. Changing sceneries, changing roles, part VI. Selected papers from the MMC Seminar, Hilversum, 16–17th May 2013 (pp. 136–144). Stockholm – Vienna – Dubai: FIAT/IFTA.

Rule, D. (2008). Professional Development: Classification Defined—Taxonomies, Ontologies and Folksonomies. Incite, 29(11), 21.

Sauer, S. (2016, March 17). User-centered design and search practices: understanding user improvisation to support creative retrieval. Presentation, ‘The Serendipity Factor: Evaluating the Affordances of Digital Environments (SEADE)’, CHIIR 2016. Chapel Hill, NC.

Schein, A. (2015, September 27). The Studs Terkel Radio Archive: Take it easy, but take it … to new places. Presentation, 46th Annual IASA Conference. Paris.

Solheim, E., and Børdalen, G. (2015, January 26). From broadcasting to publishing: creating the new foundation for NRK’s stories. Presentation, EBU Production Technology Seminar. Geneva.

Verwaest, M. (2014, June 3). The Value of Production Metadata (or the lack thereof). Presentation, EBU MDN Workshop. Geneva.

Wylie, L. (2013). RTÉ Twitter Project. In: FIAT/IFTA Media Management Commission (Eds.), Metadata as the cornerstone of digital archiving. Changing sceneries, changing roles, part VI. Selected papers from the MMC Seminar, Hilversum, 16–17th May 2013 (pp. 146–153). Stockholm – Vienna – Dubai: FIAT/IFTA.