Jeroen Bekaert sent several messages to the METS list and to me early in
May. I regret that I have not had time to respond until today and I
regret even more that my response does not fully answer Jeroen's
curiosities.
Jeroen asked: "My second question concerns the Audio Visual Schemas
proposed for use in the Library of Congress. These Schemas contains
information specific to multimedia files (audio, video, images,...). Does
there exist any relation between these schemas and the MPEG-7 (Multimedia
Content Description Interface)Schemas ? What is the main
(semantic) difference between these two ? Is there anyone who
has experience in combining the METS and MPEG-7 standard ?"
Reply: I fear that our work group has not plumbed the depth of MPEG7 and
other MPEG proposals and standards, and to some degree we have been moving
in our own direction without full study of the alternatives. In general,
we have the following impressions of the MPEG family and we will welcome
comments from those who have studied this matter more closely.
MPEG7. We understand this to be descriptive metadata, although its
inclusion of such elements as segment characterization suggests that it
can fill some of the roles played in METS by the StructMap. Our sense is
that although MPEG7 includes or can include the types of descriptive
metadata typically thought of as "bibliographic" in the library community,
the creators of MPEG7 were far more excited by the possibilities of
inscribing the "low level" characteristics of a work, the kinds of things
that are associated with knowing about pans, tilts, and cuts in motion
footage, with linguistic marks that might relate to voices and speech, and
graphic elements like colors and shapes. One of their documents (if I
understood it all) included a description of how footage over time might
be monitored to determine when a banana reached a state of ripeness.
This kind of data will be very important as tools for automatically
logging footage come into play (inputting tools), and with the emergence
of automated tools that exploit low-level data to support searching
(search tools). But as far as we can tell, these tools are not yet
mature. Since the Library of Congress is not a computer science lab and
cannot develop tools on our own, we are strongly inclined to wait for
industry-developed tools to mature before we embrace the MPEG7 standard
for, say, our video reformatting effort. (Where we play the role of
content-makers.) Out library also has an acquisition side, obtaining new
content made by others, and there we await new born-digital content that
arrives with MPEG7 data. (Haven't seen any yet.) Meanwhile, we practice
watchful waiting, knowing that our implementation of MPEG7 capabilities
will follow the availability of tools and the implementation of MPEG7 by
others. It is also possible that our implementation may still involve
METS: will we wish to wrap some new MPEG7-associated video content in a
METS wrapper? Perhaps so.
We have not studied the dictionaries associated with MPEG7 to ensure that
we have grabbed relevant terms, names of content attributes that we may
have overlooked. Since the orientation of the AV project I coordinate is
content preservation in a fairly basic sense--keeping the bitstreams
alive, worrying about transcoding--we have little immediate interest in
such things as the terminology for low-level descriptors. We are glad
someone else is working on that.
Meanwhile, our schemas are intended to do some things that I believe are
beyond MPEG's scope: (1) to record the facts about the source item we
began with--our old 2-inch videotape--and (2) to document the processes
used to copy it into a new form.
MPEG21. This proposed standard is actually more interesting to us than
MPEG7, since it is concerned with "packaging" content. I have heard it
described as "like a shipping manifest, it tells you what is in the
box." This functionality is rather like that of METS. And packaging is
especially interesting for preservation planning since it is a central
concept in the OAIS reference model for a digital repository.
Other metadata. Rather more relevant for us have been the efforts in two
American engineering societies with European connections: the Audio
Engineering Society (AES) and the Society of Motion Picture and Television
Engineers, both with links to the European Broadcasting Union. Each of
these organization has been working on metadata, albeit dividing the
territory in different ways. One AES effort is concerned (in two separate
proposed schemas) with the properties of digital audio and the facts of
reformatting processes. The working Library of Congress extension schemas
for audio and digiprov have been very much influenced by the AES effort.
We have not managed to sort through the SMPTE metadata, partly for lack of
knowledgeable analysts. We also have the impression that the SMPTE data
set is deeply concerned with the minutiae of the interior structure of a
bitstream, with content makers and broadcasters in mind. It also appears
to be an everything-in-one-bucket schema, unlike the AES examples. For
now, we don't quite know what (or how) to do something with the SMPTE
metadata. But I have the feeling that we should learn and work more
closely with the society, perhaps to try to find a middle ground for
archivists. For now, our video schema is the result of a hurry-up best
effort.
Meanwhile, regarding the characteristics of older, physical formats
(documenting what we came from in a reformatting program), we have found
the structure and codes in the Australian MAVIS collections management
database to be very thorough and impressive.
(http://www.wizardis.com.au/ie4/products/mavis/index.html). These code
lists have informed our extension schemas.
I hope this helps. I cannot tell you how tentative we feel about things,
and how much we would welcome a wise person taking a careful look at
things and guiding us to a good set of comparisons.
Best wishes --
Carl Fleischhauer
Library of Congress
|