The Detective Section: On Machine Learning and Local Knowledge Processes in A Recently Initiated Handwritten Text Recognition and Citizen Science Project
In early 2020, the research project The Detective Section: Transcription, Inventory and Visualization of Police Reports and Photographs in Gothenburg 1868–1902 was initiated by the Swedish National Archives and GPS400: Centre for Collaborative Visual Research at the University of Gothenburg. The project combines Handwritten Text Recognition (HTR) and Citizen Science methods, and the archival material consists of more than 22 000 pages of handwritten texts from the detective section of the Gothenburg police 1868–1902.
Machine Learning techniques are applied to automatically transcribe texts into data, which radically increase the usability of the archived information. To produce high quality training data for the HTR-model, and maximizing the quality of the automatically transcribed data, volunteers from the local civic society were invited to participate. By collaboratively investigating how entangled knowledge systems are produced and challenged by transcribed and visual data, the project also exposes epistemological dimensions of and in heritage research.
At the Museum Big Data Conference 2020, archivist Karl-Magnus Johansson shares his experiences from the project thus far, focusing on relationships between data and local knowledge. Professor Mats Jönsson, presents how the transcribed data will be investigated, visualized, and exhibited together with and in contrast to archived photographs, maps, and illustrations. The paper ends with a brief elaboration on the project’s general relevance for other forms of research, heritage, archive, exhibition, and knowledge production processes.
