Artificial Intelligence and Archives, 30 April 2018

Jenny Bunn Monday, 15 October 2018 19:57

Machines - like small children - may misclassify a pink monkey's tail as a flamingo Image credit Mark Bell

 






'Machines - like children - may misclassify a pink monkey's tail as a flamingo' Image Credit: Mark Bell


This event centred around two presentations; one from Mark Bell of The National Archives and the other from Olivia Vane, a PhD student from the Royal College of Arts.

Mark started things off, with a brief introduction to Artificial Intelligence, before narrowing down to machine learning; one aspect of this wider field. Using the example of his young son learning how to recognise a flamingo, he managed to convey a good grounding in the mechanisms involved and how these translated into technologies such as the use of neural networks. He then outlined how The National Archives was starting to investigate, through collaborative doctoral partnerships, the application of machine learning and similar techniques to archival problems, including predicting the applicability of certain Freedom of Information exemptions to highly sensitive data (http://www.dcs.gla.ac.uk/~graham/) and the searching and browsing of the large web archive corpus. This work, whilst still at the experimental stage, is starting to suggest that there is the potential for such technologies to support the sector in its work, particularly with regards to coping with increasing volumes of digital material.

Next up Olivia Vane discussed examples of her work in designing and implementing a variety of visualisations of cultural collections – see http://oliviavane.co.uk/for further details. Having undertaken a number of user evaluations of her work, she highlighted the issues of trust and transparency with regards to the ways in which her visualisations had been derived. In some instances, this was not felt to be a concern, e.g. in providing visualisations for explorative purposes only, but in others it was of great concern, e.g. in providing visualisations based on textual analysis of source material from which conclusions might be drawn for the purposes of research and interpretation. This prompted some discussion around whether or not archive services should be concentrating on developing their own visualisations, or on getting their collections into a form that would allow others to design and develop their own visualisations, specific to their own purposes.

Our thanks go to both the speakers and all the attendees for making this event such a thought-provoking and interesting glimpse of things that might be to come.