Demons, Dildos & Dancing Skeletons: Machine Learning at Wellcome Collection

Works held at the Wellcome Collection are being digitised rapidly, with more than 10,000 new images produced every day. All of those works are connected somehow, but making sense of that tangled old information is hard; Manual cataloguing can't add sufficient detail to every work at the same pace as the digitisation, and traditional computational tools can't capture the contextual nuances of works or the connections between them.

In this talk, Harrison Pim describe a few machine learning techniques being developed to draw meaning out of that madness, including novel approaches to search with computer vision, named entity recognition and disambiguation, and interfaces which expose the weirder corners of the collection.

Harrison Pim is a data scientist at Wellcome Collection, where he is creating tools to help visitors and researchers explore, discover, and find connections in our museum and library collection. In his work, Harrison focuses on computer vision, natural language processing and graph theory research, and turning that research into a set of real, usable services. Harrison used to work at The British Museum, and before that I did an MSci in Physics at UCL.