Media Data Science

Media Data Science includes lots of topics highly relevant to Glasgow: high performance data systems, effective multimedia archive retrieval, engagement measurement in users, recommender systems, interactive social media access to content, sensitivity analysis, ethics and transparency in algorithmic systems, voice and conversational interaction (BBC is a major early adopter and developer of Amazon’s Alexa system), subjective Bayesian optimisation, Closed-loop system analysis.

New technologies are improving the way that humans communicate (from SMS to social networks), meaning that news and comments can be shared and published to many other people, and indeed exposed outwith individual users' social circles.  What is less understood is the way this is changing and destabilising our political processes. News media are no longer the gatekeepers and moderators of information, with the immediate and visible effect of this being the democratisation of political processes through the advent of mass communication.

Theme Lead: Dr Craig Macdonald

Track record - staff

  • Ounis (politics & social media, recommender systems, topic modelling)
  • Macdonald (politics & social media, recommender systems, topic modelling)
  • McCreadie (social media and crises/emergency management)
  • Jose (Multimedia information retrieval, topic modelling, event detection in social streams)
  • Dalton (conversational interaction and music information retrieval)
  • Murray-Smith (Music information retrieval, recommender systems)
  • J.H. Williamson (Music information retrieval, visualisation of uncertainty)
  • Jensen (Music information retrieval)
  • Vinciarelli (Analysis of political discourse, Analysis of social signals in media)
  • Brewster (Interaction with Smart TVs & AR, 3D displays, in-car VR)

Projects (active in the last 6 years)

A number of previous projects included Media and Data Science research:

  • Bang & Olufsen funded multiple Ph.D. students for Brewster & Murray-Smith. One of these projects led to B&O’s Flagship product for 2015 (Beomoment)
  • Closed-loop Data Science project includes work on Recommender Systems, Music Information Retrieval, and Subjective Bayesian Optimisation of audio systems
  • Moodagent has funded multiple projects for their Music Information retrieval system with Murray-Smith & Williamson, which have led to products, and been used by 20 million users.
    • Moodagent currently funding a new Ph.D. studentship and RA (with Dalton as lead supervisor)
  • Pufferfish & Moodagent project with J.R. Williamson, J.H. Williamson & Murray-Smith
  • Co-Sound Project (Murray-Smith, Jensen, Williamson) - integrating machine learning with the archives of Denmark’s Radio (Danish BBC equivalent).
  • Interaction with the School of Gaelic to adapt Terrier to work with Gaelic (Macdonald, Murray-Smith)
  • Ounis, Macdonald & McCreadie have been involved in 2 EPSRC/Dstl projects called CROSS and ReDites on the identification of events/stories from social media and other social media analytics (EP/J020664/1 and EP/L010690/1).
  • McCreadie is co-leading the Incident Streams Project aimed at identifying actionable information during crises and sponsored by the public safety communications research division of NIST (U.S.).
  • Ounis & Macdonald have been working with political scientists at the School of Social and Political Sciences focussing on opinion leadership among users in social networks, as well as the prevalence of electorally unsound events, such as reports of voting malpractice or election-related violence.