When you hear Media Mixer, you might be thinking of a visual artist who works with different materials to create his or her works, or perhaps someone involved in audio production. You may not immediately think of the semantic web, Linked Data, or their role in making it easy to reuse and manage the copyrights for online media fragments.
Time to rethink your definition. The Media Mixer project is indeed about making the Web of Media a reality with the help of media fragment detection and semantic annotation, in conjunction with copyright management that is integrated into the Web fabric, using Linked Data principles and reasoning based on a Copyright Ontology. At the Semantic Technology & Business Conference in NYC earlier this month, Roberto Garcia Associate, Professor at Universitat de Lleida and principal investigator at MediaMixer, discussed the EU-funded effort to create, repurpose and reuse media fragments across borders on the Web, and its goal of making media more valuable for its owners such as video producers, hosters and redistributors, and more useful for consumers.
“We want to create a web-wide market for media fragment reuse,” says Garcia, who also participates in the Rhizomik initiative and the Rhizomer tool for publishing and exploring Linked Data (see our story here). Rhizomer Media Explorer is part of the proposed MediaMixer semantic architecture, described below:
Snippets of online songs, for instance, can be more easily monetized when they can be found more easily in online videos that repurpose them. But with thousands of pieces of registered content and thousands of videos on places like YouTube that might reuse them, scale is a problem. “You need a decision support system that assists a person responsible to make a claim or not in getting through all these pieces of content.”
The project takes on the tasks of fragmenting media assets, annotating them using semantic descriptions that will facilitate finding them. “It’s a mechanism to automatize the process and expose descriptions for fragment-level semantic search,” he said.
Garcia also discussed the transformation of whole videos to sets of “meaningful, indexable and reuseable video fragments,” with concept and event detection enabled via training the system with a set of annotated videos. The semantic annotation includes descriptive metadata for media characteristics, provenance metadata to credit source and specify rights, and conceptual metadata to reflect what the media is represents. In the last instance, globally unambiguous identifiers for concepts – via Web URIs from the Linked Data space – are needed, to provide more information about the concept and to enable inferencing of that concept type and relationship to other concepts. “The idea is we have access to all this data using Linked Data technologies,” he said. “The idea is we might model descriptions of particular images using semantic technologies.”
It also involves modeling licenses and policies using the Copyright Ontology, and exposing them for fragment level retrieval and re-use, including copyright reasoning. It will deliver a tool that lets people model parts of contracts or policies that had been on paper for integration into an RDF store to support reasoning about rights.
“This is trying to provide a solution for rights holders to be sure they can claim ownership and reach agreements,” Garcia said. There might, he noted, be different versions of the same song that alter copyright ownership. “You have to take all this into account before you can take a step,” he says. There also might be requirements for internal policies that must be supported, such as avoiding running a particular artist’s videos in a certain video arena. The Copyright Ontology, with its compendium of core copyright domain rights and restrictions, will drive such processes from behind the rights interface. He noted that it will be linked to schema.org in order to reuse its classes.
Says Garcia, “We use our ontology to guide all the interactions.”