MozCamp – linked media

Diagram for the LOD datasets

Image via Wikipedia - this version is in fact outdated again, by now the open resources have already doubled.

“Linked Media: Weaving non-textual content into the semantic web” – Raphaël Troncy

Traditional media cunsumption (like TV) is declining and moving to the web. The question is: how can we make media into a first class object on the web?

Lots of issues: codecs, metadata, content protection and so on. Is there a viable OSS alternative?

Media Fragments WG

Case: media fragments identification and selective retrieval of media fragments, the goal of the Media Fragments WG of the W3C. Basic principles apply: fragment identification needs to be based on the URI.

There are four dimensions which define a fragment: time (point or interval), space (rectangle for now), track (video, audio, subtitles) and id (the unique name of the fragment).

The possibilities are limited by the container format can express (e.g. quicktime and such) Protocols include http, rtsp and a lot of proprietary protocols like mms, and the various p2p protocols.

Much of the fragment identification is already possible for the most important players in the market, but the syntax is not standard in any way.

Warning, hardcore geekery ahead…

The current proposed standard uses hash marks appended to the URI, which a smart user agent has to strip off and convert into some appropriate http headers. Media servers handle the request, do the slicing and make sure that the fragments are cacheable as well.

Example: mypodcast.mp3#t=15,45

translated into the Request

GET .../mypodcast.mp3
Accept: application/mp3
Range: seconds=15-45


HTTP/1.1 206 Partial Content
Accept-ranges: bytes, seconds
Content-length: 2310034

First implementations can be based on plug-ins: Apache mod_annodex combined with for instance a Firefox add-on to create the right requests.

Media Annotations WG: Core Ontology

A couple of proprietary metadata schemas do exist, the first approach is to make the semantic meaning of all of these more explicit and to be able to map various schemas to each other.

There is a simple client read-only demo for metadata which looks very similar to the schema I used for the Twones Active API.

Another demo is about the linking of resources within the Cultural Heritage project. A very simple web interface allows for fast data entry where  terms are auto-completed with linked resources (canoncal names etc.). The interface looks like the Freebase web front-end.

The Web of Data

Interesting: the Semantic web is now being re-branded as the Web of Data. Oh well, maybe that is a good idea after all…

So what is it about:

Expose open datasets as RDF. Example DBpedia with slightly over 9M RDF triples. All of this linked to the rest of the Linked Data Cloud, which is expanding rapidly.

Reblog this post [with Zemanta]