Research & Development

What we're doing

The Data Science Research Partnership (DSRP) is an initiative between the BBC and eight partner universities. The collaboration brings together industry experts in the fields of data science and machine learning from across the BBC and world-leading UK data scientists from the Universities of Bristol, Manchester, Edinburgh and Surrey, Imperial College London, Queen Mary University of London, Ulster University and University College London.

The aim is to create a body of research, insights and prototypes that can make a real impact on the BBC and its audiences.

There are four main themes of work:

Understanding audiences: Use data to understand better what our audience wants from the BBC, why they want it, and what impact these programmes or services have on them.

Understanding content: Explore what machine learning can teach the BBC about its programmes and services and what it stands to gain from it.

Curation and personalisation: Create a more personal BBC, designing tools and algorithms to help programme-makers with editorial and commissioning decisions.

Content of the future: Design new and innovative audience experiences.

A critical part of the DSRP is developing data science skills to address the scarcity of data scientists in the UK. Working with our network of university partners, we promote work-based learning opportunities for students at undergraduate, postgraduate and PhD levels to ground their learning in real-world problems.

The DSRP is curating a selection of datasets for use by the research community. These datasets will be built from the BBC’s data and released to the community with the hope of stimulating research.

Why it matters

Within the BBC, we see the huge potential for more data-informed experiences and tools. The BBC and universities share a keenness to exploit opportunities in data science, and there is a huge appetite to work on the BBC’s problems and data sets, with a shared focus on the public good.

The DSRP is helping colleagues in the BBC develop the skills they’ll need in a data-driven future.

Outcomes

Core to the DSRP is the creation and oversight of collaborative projects between the BBC and academic partner institutions. “To date, the DSRP has enabled over 40 such collaborative projects, and significantly, supported PhD students, finding placements for secondments and interns. Schemes such as these are of great benefit for both the BBC and academic institutions as they encourage the sharing of knowledge and experience.

As well as initiating and running collaborative projects, the DSRP supports broader initiatives to benefit the research community. Examples include AHRC Creative Clusters and supporting a Prosperity Partnership to develop and trial new ways to create and deliver object-based media at scale. These partnerships are business-led research partnerships between leading UK-based businesses and their long-term strategic university partners, with match funding from EPSRC, part of UK Research and Innovation. We have hosted several fellowships. “Vectors and tensors for textual understanding of subtitles and news” was sponsored by the Royal Academy of Engineering and “MTStretch”, a fellowship focussed on machine translation of low resourced languages was sponsored by EPSRC.

The DSRP has helped shape three Centres for Doctoral Training in Data Science: QMUL’s Artificial Intelligence and Music (AIM) and Data Centric Engineering (DCE); and UCL’s data-intensive science CDTs.

PhD projects

Title University
Merging Augmented Reality with Television to Create Novel Media Experiences University of Manchester
   
Data-Intensive Tools for Creative Practitioners University of Manchester
   
Automatic Extraction of Rich Metadata from Broadcast Speech University of Edinburgh
   
Measuring audience immersion in visual media University of Bristol
   
Multimedia Computing with Interpretable Machine Learning Queen Mary University London
   
Learned video processing for high quality content delivery Queen Mary University London
   
Intelligibility through design in AI and data-driven media University of Edinburgh
   
Generating virtual camera views using Generative Networks University of Surrey
   
Building Journalist Knowledge Graphs for Exploratory Text Analysis University of Manchester
   
Multi-modal content similarity for predicting audience behaviour University College London
   
Computational Lighting using Generative Networks University of Surrey
   
Semantic understanding of TV programmes to enable automatic length adjustment or extraction of stand-alone portions Queen Mary University London
   
Data analytics for video distribution applications Queen Mary University London
   

Internships, secondments and fellowships

The DSRP enthusiastically supports the cross-fertilisation of skills between the BBC and academia via internships or secondments. The movement of staff and students between the BBC to a university research group and vice versa has huge benefits for both institutions, exposing the intern or secondee to new ideas and allowing the development of new skills.

BBC R&D - Explainability for Domain Experts

BBC R&D - Using artificial intelligence to automatically segment media content

The DSRP has been hugely successful in finding placements for interns and secondments within the BBC.

Title University Type
Vectors and tensors for textual understanding of subtitles and news QMUL/UCL Fellowship
     
Understanding, analysis and visualisations from BBC online taster pilots University of Manchester Internship
     
Experimenting with User Experience Object based media data – CAKE University of Manchester Internship
     
Reinventing the cost function for World Service Onward Journeys University of Manchester Internship
     
Singing with Machines QMUL Internship
     
My data rights - your data obligations QMUL Internship
     
Measuring the Immersivity of UHD and HDR Television University of Bristol Secondment
     
Machine Comprehension for News QMUL Internship
     
Measuring the Societal Impact of the BBC University of Ulster Internship
     
Aligning BBC Algorithms: Editorial Guidelines, good practice and the law UCL Internship
     
A Hybrid Recommender System for Text Content University of Manchester Internship
     
Camera Correspondence and Scene Understanding University of Manchester Internship
     
Intelligible Cloud and Edge AI (ICE-AI) University of Edinburgh Secondment
     
Graph-based recommender methods UCL Internship
     

Future

The DSRP continues to search for new research areas and encourages as broad a range of partners as possible. The partnership aims to collaborate with media and technology organisations from across the UK and internationally on a range of projects. We will continue to share more data and problems with our partners. This partnership will train the next generation of data scientists on media problems that will create audience experiences that don't even exist yet.

For more information

Andy Secker, DSRP lead, is the primary contact for enquiries.

The DSRP is part of the work of BBC R&D's Data Team, led by Andrew McParland and supported by the Partnerships lead, Sam Chadwick.

Tweet This - Share on Facebook

BBC R&D - Partnerships

University of Bristol

University of Manchester - Data Science Institute

University of Edinburgh - Data Science initiative

University of Surrey

Imperial College London - Data Science Institute

Queen Mary University of London

Ulster University - Data Science

University College London - Big Data Institute

Engadget - The BBC is turning to AI to improve its programming

IT Pro - BBC to develop new services based on machine learning

The Verge - BBC will use machine learning to cater to what audiences want to watch

Advanced Television - BBC seeks to unlock data potential

TVB Europe - BBC plans five-year project to 'unlock potential of data'

Topics