There’s a growing focus on the opportunity for semantic technology to help out with managing media assets – and with making money off of them, too. Last week, The Semantic Web Blog covered the EU-funded project Media Mixer for repurposing and reusing media fragments across borders on the Web. Also hailing from Europe – France, to be exact – is Perfect Memory, which aims to support the content management, automatic indexing and asset monetizing of large-scale multimedia.
Perfect Memory, which was a finalist at this spring’s SemTechBiz semantic start-up competition, has implemented its platform at Belgian TV broadcaster and radio RTBF, for its GEMS semantic-based multimedia browser prototype that was a runner-up at IBC 2013 this fall. In September it also received a 600K Euro investment from SOFIMAC Partners to extend its development efforts, platform, and market segments, as well as protect its innovations with patent filings.
“Our idea is to reinvent media asset management systems,” says Steny Solitude, CEO of Perfect Memory.
Organizations like RTBF have massive content but can’t get optimal value from it using traditional relational database systems, he says. And it’s not just traditional media companies that are experiencing these pangs. “Red Bull is a TV Channel today,” he says. “It has lots of content. Every kind of brand or trade today wants to be a media content company.”
To do that right, he believes, requires deploying a system that automatically will manage content indexing and enrichment, which costs less than if you were to manually take on the task, and turn that content into linked knowledge by connecting any asset to its pertinent associated Linked Open Data.
Perfect Memory’s semantic middleware, available as a platform as a service, works with companies’ existing infrastructures, generating from them a triple store and associating all the semantic triplets to a knowledge base and to a service that ensures the mapping of the content. It serves, he explains, as “one unique search point to all your existing asset management systems, because we sit on top of that,” converting external sources into semantic-understandable content. “And, as that is available in our semantic middleware, any kind of process—for example, content analysis or media processing – can access this source now.”
The platform, Solitude says, is the first of its kind to be compliant with the EBUCore metadata specification, which is based on Dublin Core and was designed to offer a list of attributes to describe audio and video resources for broadcasting applications. “That is the way that you have to structure and index audiovisual and multimedia content if you want it to be flexible and available anywhere,” he notes.
Running a process like content enrichment for assets now covered by its semantic middleware – such that you can identify named entities and make connections with other knowledge available on the Linked Open Data cloud – adds up to “liberating the exposure of this content,” Solitude says. A large part of information to enrich assets already is available on the semantic web, he says, and organizations have to develop a strong way to build knowledge from available outside knowledge sources.
“If you have media on Barack Obama, if you make that media enriched, you can associate it to simple videos of things like Michele Obama, president of the U.S., Washington and so one,” says Solitude. “So, if you make a search using one of these different entry points you will be able to find Barack Obama. So you multiply the possibility of finding and exploiting your content, and that’s why indexing and enrichment are so important.”
The more that content is strongly and precisely structured and prescribed, the better that content can be exploited, and the better the odds are of monetizing it. With semantic and rich annotations in place, organizations can associate materials to messaging or advertising opportunities.
“Often when you hear about Big Data, people mean text. But Big Data is multimedia and one difficulty with that is you have the time [element], and it is very hard to cope with time-coded information and know how to deal with chapters of content and structurization, and the fact that this part of the video can be found in another video or edited in another content,” he says. Perfect Memory supports these requirements, with the EBUCore spec as well as the Open Open Archival Information System for prescribely describing the structure of information to be exchanged at its heart.
Says Solitude of Perfect Memory’s expectations: “We expect to be one of the leaders associated to Web 3.0”