researcher working on "future of TV" EU project, NoTube,
VU University Amsterdam
Social media is becoming more prominent each day. However, users perceive social media as just another way to communicate or as a fancy tool for communication professionals. Dan Brickley proves social media has a lot more potential. Brickley and his colleagues are integrating social media with television in the NoTube project. They use Hadoop to explore the social web data they need.
Brickley works on the NoTube project at VU University Amsterdam. NoTube integrates television and the web: “In particular we look at social uses of the web in relation to TV such as debate and collaboration. Or interoperability between phones and TVs, independent from their software systems. For example we have been building hybrid TV recommendation systems that use Linked Data information from the web that describes TV content, viewers, their interests and activities.” Besides applications for viewer interactivity, Brickley is also researching possibilities for futuristic remote controls: “We have been exploring the future of remote controls, using smartphones and tablet PCs. Such devices could help make TV more personal and social.”
Of course many companies are also exploring these ideas. The difference is that NoTube is a publicly funded initiative. This means the project can focus on user needs: “Most current products only allow phones and TVs to communicate if they use the same software and hardware. NoTube makes TV content a focus for communication, sharing and discussion, regardless of people’s purchasing decisions.”
To integrate television and the web, Brickley uses Hadoop to explore data from social media: “For example, if I listen a lot to jazz music on Last.fm, or if I’ve bought every book by Stephen Fry via Amazon.com, these facts can be used to create a very personalized TV experience. Our early experiments confirmed that this could provide a better experience than having an interactive TV ask new users everything about themselves. However, we also found that this kind of data was quite sparse, and its usefulness varied a lot from person to person. This is why we began investigating the use of larger background datasets: not just Twitter data from one viewer, but from every viewer.”
To investigate these large data sets, Brickley has used SURFsara’s experimental Hadoop cluster: “Using some large-scale Twitter and Wikipedia datasets, we ask questions regarding a massive collection of many users’ data: What kinds of people follow Stephen Fry on Twitter? What sort of news links do they share? How do the celebrities people follow on Twitter relate to the kinds of media content they’re interested in? Without software like Apache Hadoop, this kind of analysis is difficult to undertake due to the scale of the data.”
Brickley is enthusiastic about the experimental cluster: “It has made clear that Hadoop is a very powerful tool to explore recommendation strategies that work with web-scale data. The most impressive aspect of these explorations is how SURFsara’s facilities and support, combined with Apache’s open-source tools, hides complex details of distributed, ‘big data’ computing. This allows us to focus on our problem domain rather than on the computing platform.”