Here's a summary of what we got up to for the period 6-17th October, 2014.
Data Team
CODAM
The CODAM project around video fingerprinting and visual search, funded by InnovateUK has officially started with a kick-off meeting with all partners and an InnovateUK representative attending. We also made good headway on collecting use-cases across the ±«Óătv, one of our first work packages, as well as collecting test data. All joint effort by the data team, especially Andy talking to media managers from all over the ±«Óătv.
COMMA
The COMMA team looked into integrating search into the system and have reviewed new approaches to accounting and account management components. They also looked at work on centralized logging and monitoring of a running cluster for debugging and integration purposes.
Discovery Team
The discovery team has been running a small scale, closed audience test of a prototype app, trying to understand the expectations and attitudes towards personalised curation of content. They took over a couple of walls at the office for experience mapping, and used that as a good excuse to plaster them with hundreds of post-its.
Devices Team
Mediascape
Chris Needham, Sean and Libby were in Munich for a MediaScape face to face meeting this week. At the meeting Libby encouraged the partners to document their reasons for making prototypes and the audiences they had in mind, based on our current .
Radiodan
Lots of prep for our upcoming radio home trial this fortnight including a tiny update system, cleverly called “updater” that downloads the latest software onto radios meaning we can deploy buy fixes out in the field. Well, not in the field, but in our homes.
This week Andrew N also added gratuitous transitions to our Physical UI (buttons, lights etc). It’s amazing what a difference to the experience smooth transitions between colours makes. Part of this is support for easing functions - there are loads of good tools online for and them.
Recommendations
Fabian, our visiting MSc student from the University of Zurich and Chris Newell have been exploring the relationship between recommender system accuracy and diversity. Accuracy in this context is the ability of a recommender system to predict future viewings, which is commonly used to measure recommender system performance. Fabian has implemented some additional metrics for the which measure Intra-list Diversity (the diversity of genres seen by each individual user), Overall Diversity (the diversity of content seen by all users) and Personalisation (how different users' recommendations are from each other). Ideally, a recommender system would maintain accuracy whilst maximising diversity and personalisation.