Introduction
One of the pre-requisites for trading "virtual goods" is that everybody in the value chain must know exactly what they are talking about. Without agreement on meaning, it will be extremely difficult, if not impossible, to make deals and transact business between parties who do not know each other. And as we are talking about content that may be compiled from several different sources and contain several different media types, such as music, text and video (i.e. true multimedia), we are potentially talking about hundreds of different metadata and identification systems. When talking about music we use ISRCs, ISWCs, GRids, MWLIs, IPIs and ID3. For textual resources we have ISBN, ISSN, ISTC, BICI, SICI, NITF, PRISM and ONIX, for visual content there are ISAN, V-ISAN, UMID, MPEG-7, DMCS and SMEF. Museums, libraries and Universities have their own systems (independent from the schemes based on content-types): IIM, LOM, IMS, CIDOC and MARC. To finish the alphabet soup for this paper, there are identifiers for physical products (EAN and UPC) as well as identifiers and metadata for the online world (DOI, DII, URL, URI, URN and iDD).

When trading a piece of multimedia content it will be necessary to be able to deal with identifiers and descriptors from any these identifier and metadata systems. The alternative – the development of an entirely new unitary system that would be adopted by everyone – is highly attractive, but probably completely impossible on the basis that everyone would have to agree on the new system and to give up their own systems.

In other words, we will need to create some method to be able to map the semantics of one metadata standard to the semantics of another if we want to be able to create true multimedia experiences.

Lost in translation
Nowhere is this more obvious then when dealing with content that is governed by rules articulated in rights expression languages (RELs), such as specified by OMA (2004) and MPEG (ISO/IEC 21000-5:2004). These two standardisation bodies each opted to adopt a different rights expression language on the basis that they were the most appropriate for their respective domains. It causes, however, a potential problem for users (a user being any participant in the content value chain, from content creator via content distributor to the consumer), namely that content that has been created in, say, the "MPEG domain" and is governed by the MPEG REL, cannot be rendered by a device in the "OMA domain" which does not understand the MPEG REL, even though the underlying compression and packaging of the content is the same.

The film "Lost in Translation" we all saw on the silver screen last year showed that translating from one language into another can be tricky. While this is true for humans, it is even more so for computers – especially when commercial values are at stake – as the following anecdote indicates. During the discussions between Consumer Electronics (CE) industry engineers and executives from movie industries which led to the drafting of the MPEG REL specification, there was extensive discussion about "deleting" content.

However, while the CE engineers understood "deleting" as the process of wiping the entry from the media’s table of content (and thus making the file inaccessible), the content owners’ view was that “deleting” should mean a complete overwrite of the entire file with random numbers, thus destroying every trace of it and making it completely impossible to restore. This seems to be a small difference, but there are significant consequences as, if devices were to be built based on the former definition, content owners might well not have been willing to release their content for such devices. This story highlights the critical importance of well-defined and agreed semantics.

Managing Meaning
Douglas Adams has already described a solution to this problem: A "Babelfish" (Adams 1979) that translates entire sentences without any loss of meaning from one language into another. Unfortunately, no-one has been able to implement a complete Babelfish as of yet.

We do, however, have plenty of syntactical tools (XML being the fashion of the last couple of years) to help us with the transforming the structural grammar and we have many online dictionaries that can help to translate individual words. But when it comes to translating phrases or sentences, the available systems are far from perfect. What has been missing up to now are semantic tools that can translate (i) from one language to another language without losing the meaning, but also (ii) to translate from one environment to another environment (e.g. between different content verticals as discussed above) while maintaining the meaning of what is being translated.

However, with the development of the MPEG Rights Data Dictionary (ISO/IEC 21000-6: 2004) as part of the MPEG-21 group of specifications we do now see tools emerging that should be able to solve the semantic interoperability problem.

MPEG Rights Data Dictionary Approach
When MPEG set out its requirements for a rights expression language and a rights data dictionary it was not anticipated that one of the submissions would be an ambitious and novel idea for the creation of tools for semantic interoperability. The submission from the Contecs:DD consortium (at that time: International DOI Foundation, Melodies and Memories Global Ltd., the Motion Picture Association, the Recording Industry Association of America and Enpia, who have since then been replaced by Rightscom Ltd.) was chosen by MPEG partly because it did offer a route to interoperability, enabling MPEG to work with the huge variety of vocabularies implied by the profusion of metadata schemes identified earlier.

The rationale for this decision was that communities wishing to use MPEG technology would not necessarily want to adopt a single (new) MPEG vocabulary, but would continue to use their own. Indeed, it is not the job of a horizontal standards organisation like MPEG to dictate to specific vertical communities what they should and should not do within their own sector. This of course extends to enabling them to continue to use their own metadata schemes, even though the use of a single scheme could greatly enhance meaningful communication between sectors. It was this problem that MPEG sought to solve when it adopted the approach of an ontology-based rights data dictionary. This means that the dictionary is built up as a knowledge base using a consistent data model with all terms being expressed in terms of their relationships to one another. For computational purposes this enables extensive inferencing, which both cuts down complexity and achieves rich results from the knowledge base (cf. International DOI Foundation 2004).

The dictionary standard is actually based on a remarkably simple model, containing only four entities – resource, agent, time and place. Combining these four entities in a "Context Model" (so called because each term is analysed in terms of the context in which it exists), it is possible to classify and derive terms for the dictionary in a highly granular way with the use of these four entities. The advantage of using an underlying data model of this nature is that the dictionary can be cumulatively enlarged in a consistent manner, so that all terms are potentially interoperable, even though they come from non-interoperable sources. For more information see International DOI Foundation (2004).

This is achieved by analysing each term as it is presented for inclusion in the dictionary, then mapping it to a central core in accordance with their original semantic content. By this means, the dictionary can be built up with terms from many different vocabularies, mapped together in a matrix of meaning.

The dictionary as finalised and published in the ISO standard is only small, but, supported by the Context Model it contains the building blocks of a potentially much bigger dictionary. And given that the communities that may adopt MPEG standards could be very substantial, this bigger dictionary will contain terms required by anyone wishing to use MPEG technology, especially, but not limited to, the MPEG Rights Expression Language (ISO/IEC 21000-5). The process for extending the dictionary is the proposed Registration Authority, which is expected to be managed by the International DOI Foundation (IDF). This is significant because the IDF represents a major content owning community that will be encouraged to adopt the dictionary from the start. In addition the music, motion picture and publishing industries have all expressed their support and several implementations are currently under way.

Achieving semantic interoperability between MPEG and OMA
While the dictionary deals with the method by which terms can be made interoperable, it remains to be seen how rights languages themselves could be made to interoperate. To understand this, it’s essential to understand the problem that multiple rights expression languages may present. Say that rights holder A uses the MPEG rights expression language while rights holder B uses the OMA language. Both languages have a right called "play". The question then arises as to whether the MPEG "play" is the same as the OMA "play". But only by analysing the semantic content of both versions of the word "play" it is possible to know if they mean the same. If they do not mean exactly the same, there is a danger that a device will allow a user to deal differently with a resource, depending on whether the device is using the MPEG "play" or the OMA "play". This could have serious unintended consequences and may lead to the same issues as discussed above with respect to "delete".

One solution to this is to use an interoperable rights data dictionary, constructed on the MPEG principle, to enable users to generate rights expressions in both the MPEG and OMA languages, by using the same top-level core terms, which are then translated ("specialised", to use the term coined in ISO/IEC 21000-6) into the appropriate MPEG and OMA semantics. This approach would ensure that the actions permitted by an MPEG or OMA rights expression (using terms from the respective languages) were equivalent even though the two rights expressions had apparently different terms. There may, of course, be other methods to achieve the same ends, but what is certain is that direct translation between rights languages may be unreliable for a combination of syntactic and semantic reasons. If this is so, a better solution may well be the one outlined above.

What does that mean practically?
Rights owners will describe their content as well as the rules under which their content can be accessed in their preferred language and to their requirements. Device manufacturers A and B will, however, design their devices with technical capabilities in mind.

This will almost certainly lead to slightly different implementations of, say, the verb "play": In a specific device, "play" could involve a resizing of a video clip to a slightly smaller screen and another device it would involve the reduction of the colour depth to a black-and-white picture to cater for a black-and-white display.

In order for the an automated content distribution system to work with such different devices, a semantic connection between the content owner’s "play" and the two device’s "play" needs to be created so that (i) the former becomes a superset of the two latter and (ii) that this relationship becomes known to the content distribution system as well as the devices.

Bottom line
Everyone agrees that standards are valuable and can lead to interoperability. But when there are different standards solving the same problem in different domains, it may be extremely difficult to efficiently interconnect even adjacent domains. Today we have this situation: each content vertical and each distribution domain has its own vocabulary – with the net result that true multimedia must remain a dream unless there is a process to make controlled vocabularies interoperable.

Technologies such as the MPEG-21 Rights Data Dictionary can help to manage these various sets of meaning so that one always knows in terms of one’s own vocabulary what someone else was saying.

Sources

  • Adams, Douglas (1979): The Hitchhiker's Guide to the Galaxy. New York: Harmony Books 1979
  • International DOI Foundation (2004): DOI and data dictionaries (Version 2.1). Available from http://www.doi.org/factsheets/040707DOIDataDict2-1.pdf
  • ISO/IEC 21000-5:2004. Information Technology – Multimedia Framework (MPEG-21) – Part 5: Rights Expression Language (REL)
  • ISO/IEC 21000-5:2004. Information Technology – Multimedia Framework (MPEG-21) – Part 6: Rights Data Dictionary (RDD)
  • Open Mobile Alliance (2004): OMA Rights Expression Language. Candidate Enabler Release V2.0. Document OMA-DRM-V2_0-20040715-C

About the authors
Niels Rump has worked in the area of DRM since the mid 1990s. He was the main developer of one of the earlier commercial DRM systems (Fraunhofer IIS' Multimedia Protection Protocol, MMP). During his time at Fraunhofer IIS, he started working in several DRM-related standards bodies including MPEG, AES, and SDMI. He has also worked for InterTrust Technologies before joining Rightscom in 2001 where he concentrates on the technical aspects of DRM applications and technologies. As a Senior Consultant he is involved in the development of identification, metadata and messaging systems within, amongst others, the Music Industry Integrated Identifier Project (MI3P). Niels is a regular speaker at conferences on DRM and has published several papers on the topic. He holds a degree in computer science from Erlangen University, Germany. Contact: niels.rump@rightscom.com

Chris Barlas has more than twenty years experience of rights management. In the mid 1990s, he led the European Commission supported Imprimatur project. Subsequently he was involved in other successful European Commission projects including which delivered the widely adopted analysis of metadata interoperability. He has also worked as a writer and producer in television and radio. As a Senior Consultant at Rightscom, he has advised a leading software company on its eBooks strategy, a major distance learning institution on third party rights management and an international bank on its publishing work flow technology. In the public sector, he edited the CEN/ISSS DRM study and co-authored WIPO's recent report on DRM. Chris has been active in international standards development. At MPEG, he co-edited the MPEG-21 Rights Data Dictionary, published in April 2004 and took an early leadership role on standards at the Open eBook Forum. At Rightscom he recently assumed responsibility for developing the market for Ontologyx.

Status: first posted 10.01.2005; included in INDICARE Monitor, Vol. 1, No 8, 28 January 2005; licensed under Creative Commons

URL: http://www.indicare.org/tiki-read_article.php?articleId=68