The Congruence Engine project
The BFI has recently completed its involvement in the AHRC-funded and Science Museum-led Congruence Engine project which sought to bring together collections in industrial history, using new digital techniques, under the headings of Energy, Textiles and Communications. Because of Britain’s rich history in industrial filmmaking, the Archive holds crucial collections under all three headings. Curators Patrick Russell, Ros Cranston and Steve Foxon, and Head of Data & Digital Preservation Stephen McConnachie all contributed expertise – and data – to the project. In addition to supplying data and information on our collections under the three headings, the BFI led a close investigation into (selected episodes of) the Mining Review series which it preserves in its entirety (some 450 issues all told). For these episodes, the films themselves, the Special Collections papers documenting their production processes and previous BFI-authored writing were subjected to machine learning experiments.
Experiment one involved creating a pipeline to join together a few machine learning solutions for enhancing the metadata of Mining Review collections:
- Speech-to-text using various models derived from Whisper to generate text outputs
- Natural Language Processing powerhouse spaCy to undertake Named Entity Recognition on those text outputs
- Wikification tool Entity-Fishing to resolve those entities to Wikidata
Experiment two involved testing the video understanding capabilities of Google Gemini 1.5 Pro to create metadata for a Mining Review episode using examples from the BFI National Archive’s expert human cataloguers as templates – including:
- Detailed shot-by-shot description aka shotlisting
- Content summary with attention to coal mining as the focus
- Article-writing, placing the episode in the context of the Mining Review series
These were then reviewed from a curatorial perspective. A chapter in the project book co-authored by Stephen, Patrick and Congruence Engine staff will summarise this research package and its implications. The edition of the series on which we particularly focused is Mining Review 10th Year No. 11 which can be viewed on Player here – check it out, it’s a cracker!
The Congruence Engine project final report has been published, and project lead Dr Tim Boon summarised the project at the Towards a National Collection closing event.
– Patrick Russell, Senior Curator (Non Fiction) & Stephen McConnachie, Head of Data and Digital Preservation
Archiving in the Age of (Super) Abundance livestream
The advent of online video in the early 1990s was revolutionary, causing seismic shifts in film content, production and consumption. Its arrival and subsequent development have permanently altered the nature of moving image itself, and its relationship to society.
Along with this come the challenges of collecting and preserving this material, and ensuring it remains accessible for future generations. Online moving image content remains critically underrepresented in the UK’s film and moving image archives. With support from the National Lottery, the BFI National Archive is addressing these challenges through an accelerated acquisition programme, alongside the research and establishment of the infrastructure that will enable the on-going collection of such material.
To support research on the project, we’re hosting a one-day symposium at BFI Southbank that explores the history, development and current landscape of online moving image, and the new opportunities and challenges it presents for creators, audiences and archivists.
Tickets for the in-person event quickly sold-out but, due to popular demand, we’ve arranged to livestream the event. Find out more and sign-up to be one of our digital delegates here.
– Alex Prideaux, Marketing and Events Manager (Our Screen Heritage)
The Inside the Archive blog is supported by the BFI Screen Heritage Fund, awarding National Lottery funding.
