The Past, Present, and Future of Metadata
When asked how to handle metadata that is 'rubbish' or misleading, Jackson says the trick is to wrap a clip it in multiple lists, based on the idea that one of the metadata lists will be accurate for any given point of video. This is consistent with earlier ideas of multiple passes on audio and aggregation of real-time and post-event close captioning, as false positives are less likely in metadata searches than complete inaccuracies.
Towards A Unified Standard
A final question from the audience set off a long discussion on metadata standards, specifically whether MPEG-7 or RDFa is taking hold?
Jackson says RDFa is at the core of his project's tools, but solely as an internal decision.
"We haven't gotten around to RDFa but we'll probably go to MPEG-7," said Jackson. "Right now Media RSS and HTML are what we see on several systems. We're surprised at how semantic the metadata is, so we can do an XML parsing on sites like YouTube and have relatively consistent results."
Waitelonis says they have implemented MPEG-7 but are also working on RDFa to publish content to a semantic web.
"I think RDFa is important and may eclipse MPEG-7," he said, but was quickly countered by Berry.
"Studios are currently only using attribute metadata," said Berry, "but they are starting to graduate to temporal metadata; we expect they'll start to proliferate time-based metadata in MPEG-7."
An audience member, who said she was part of the original MPEG-7 specification, reminded the group that MPEG-7 time-based metadata could be integrated into an MPEG-4 system container.
"The Media Annotation working group in W3C is looking to address conversion of metadata from MPEG-7 to other metadata tools," she said. "They are trying to answer the question as to how to handle extensibility with compatibility."
Another audience member noted that Transmission.cc has an ongoing discussion around a metadata standard, with a draft standard already out that discusses the issues with vocabulary and suggests a way around the vocabulary limitations.When asked about the issue of converting metadata from one format to another, Waitelonis says that he hasn't currently facing the issue of converting metadata from one framework to another since many who use his metadata are only extracting content from the database, rather than trying to merge it with other metadata databases.
"MPEG-7 is no worse than PBCore, or Dublin Core, or even Media RSS, although the latter isn't supposed to be extensible," Jackson added. "Media RSS became a bit of a standard because a major player created a large block of content."
"It's the wild, wild West," Berry agreed, noting that "everyone may claim to use the de facto standard of Media RSS, but they also have their own extensible versions."
"Right now there is a big legality on who owns the metadata," said Berry. "Metadata is currently not being done at production level, but it very possibly could be done that way at some point. Today, though, it's a very expensive proposition for the studios to create the metadata, so they will tend to hold on to that content and not push it out on to the Net."
Related Articles
User interface optimized to let users drill down through customized time slices
21 Jun 2010