Understanding Multimodal Data at Scale

Edinburgh, 30th November-1st December 2015

Organisers: Steve Renals, Maja Maricevic, Bill Byrne

The workshop will focus on large scale processing of text, audio, and video data, which presents the scientific challenges of recognition, generation, translation, understanding, and knowledge extraction from heterogeneous data sources, at a large scale. This interdisciplinary workshop will involve a number of research disciplines necessary to meet these challenges: machine learning, natural language processing, computer vision, speech & audio processing, and digital humanities. In addition to including a broad range of researchers across these areas, the proposed workshop will also include non-academic stakeholders from libraries and the BBC. The scope of data to be considered is thus broad, including broadcast and internet media, social media, and large historical collections, as well as informally collected data from smartphones and the like.

The proposed workshop is timely both in terms of recent scientific advances (e.g. deep learning), combined with the availability of compute resources, and user requirements across a number of application sectors.  The UK’s activities across the component areas are extremely strong, and an important aim of the meeting is to develop an interdisciplinary research community in which stakeholder challenges can drive fundamental research, which in turn can address these social and economic challenges.

The workshop will focus on:

  • Developing an interdisciplinary research community
  • Integrating foundational and translational research
  • Defining the key research questions that a scientific programme in the area should address, and the key applications and challenges that would drive such a programme
  • Scoping a programme of activity in the area, with targets for the shorter and longer term, and making clear the added value of doing such research in the framework of the Alan Turing Institute