BBC R&D

Posted by Matt Haynes on , last updated

A week of new collaborations and outreach for IRFS, also the COMMA project begins to float in the "cloud" and our automated clustering of programme scenes gets some new data.

Collaborations

This week we've been busy again with our external collaborations and outreach.

Firstly we welcomed Michael Barroco to our engineering team, he has joined us for a 6 month attachment from the EBU and brings his expertise and interests in distributed, realtime and scalable systems.

Chris Needham has been working with some undergraduate students from our shared office space with UCL, we are working together on our HTML audio waveform visualisations

Tristan has been over at City University giving a lecture to MSc students on crowdsourcing and the World Service archive project.

Michael, Sean and Yves partnered with Sidekick studios at the Radio & Music Connected Studio event, they pitched a prototype to make news appealing to a young audience. 

COMMA

Myself, James and Yves have been working towards our trial run of the COMMA platform, which will perform speech-to-text transcription on a dataset donated by the Cheltenham Literary Festival.

For a long time our engineering team has been using automated builds with continuous integration and deployment practices, but our use of Amazon Web Services for the COMMA project allows us to take this one step further. This week we have set up a completely automated build of our entire infrastructure, allowing us to drop and recreate the whole stack with the click of a button.

Clustering Scenes

The Content Team has been working on comparing methods to best cluster scenes within TV programmes, this is part of their exploration around chapterisation of programmes to allow surfacing of interesting segments.

Barbara and Zillah have been manually annotating programmes, Anthony and Thomas have been extracting data from tweets to use and Denise is continuing to work on her machine learning suite to automate extraction.

Recommendations 

Chris Newell has been exploring how we could use Redis (a high performance key-value store) to process audience data in applications such as collaborative filtering. Initial results look very promising. 

MediaScape

Libby has been arranging and preparing for the MediaScape user scenarios workshop next week, with help from Jo Larkin, and also Jo, Andrew W and Theo.

Chris Needham has been working on a description of overall concepts as input to the architecture work in the project.

FI Content

Barbara has been continuing her work on FI Content in relation to the Open Call. The project partnership is working on the job of identifying and selecting external experts to act as reviewers appointed by the EC to select the winning proposals.

Chris G and Theo have been talking to BBC iPlayer about the project work we've been doing with Lancaster University on FI Content.

Speakerthon

Zillah, Michael and Tim have been working on the UX requirements for the upcoming Speakerthon event. 

Interesting Links