The Congruence Engine: Digital Tools for New Collections-Based Industrial Histories

Lead Research Organisation: Science Museum Group
Department Name: Science Museum Research

Abstract

The capacity to make strong connections between historical objects and sources lies at the heart of this project as it does in the everyday museum and historical practices that it is designed to support. Curators creating displays combine artefacts, images, audio-visual materials and histories. Family and local historians connect records of ancestors and localities to establish their genealogy or to understand the past of where they live. Academic historians patiently and critically connect a diverse range of archive sources with existing literature to tell new stories about the past. All rely on connecting different fragments of the past as they create the tapestries of narrative that constitute our local and national histories. The Congruence Engine will create the prototype of a digital toolbox for everyone fascinated by the past to connect an unprecedented range of items from the nation's collection to tell the stories about our industrial past that they want to tell. Until now, we have become acclimatised to a world of research where it has only been possible to work with a selection of the potentially relevant historical source material for any historical investigation we want to undertake. And now, in our information society, we expect to go to a search engine and find a record of anything. But so often such searches disappoint, and for two main reasons. First because the tyranny of the free-text search where ranked results lists favour the results of previous searches, and cannot be guaranteed to include the full set of what is relevant to the search. The second reason is that the records of so very many of our heritage collections are thin, inconsistent, or kept in institutional siloes hidden from outside access. This project explicitly works with these collections that are generally represented by weak data. In place of the two-dimensional ranked list of search engines, we aim, with 'The Congruence Engine', to model a world in which users will be able to explore data neighbourhoods (technically 'knowledge graphs') where a great diversity of information about heritage items that are deeply relevant to their investigations will be readily to hand - museum objects, archive documents, pictures, films, buildings, and the records of previous investigations and relevant activity.
Building on the successful experimentation of 'Heritage Connector' (the Science Museum's TaNC foundation project), this major project will develop a repertoire of prototype discovery tools to access the industrial and related collections brought into the study from our investigating and collaborating organisations and partners. To achieve this breakthrough in collections accessibility, it will bring together in collaboration a unique combination of skills and interests. Here, digital researchers will work with professional and community historians and curators to address real-world historical investigations of Britain's industrial past. Through 27 months of iterative exploration of three industrial sectors - textiles, energy and communications - the digital researchers will work with the historians and curators, tuning the software to make it responsive to user needs. They will responsively use computational and artificial intelligence techniques - including machine learning and natural language processing (specifically, eg, named entity recognition) and a suite of bespoke entity-linking routines - to create and refine datasets, provide routes between records and digital objects such as scans and photographs, and create the tools by which the participants - who will not need to be digital experts - will be able to enjoy and employ the sources that are opened to them in the construction of narratives. These narratives will be expressed in the project's mobile digital exhibition space, on its website and a variety of conventional popular and academic outputs. Software will be made available via GitHub; we will produce 'how to' guides.

Publications

10 25 50