HTR for Serial Sources: the 1936 Genoa Census as a Case Study

250,000 handwritten household record sheets from the 1936 Genoa census have been inaccessible for decades — no name index, no searchable text. This PhD project uses Transkribus to change that, revealing both the opportunities and the limits of HTR for serial sources: extreme variation in handwriting poses real challenges, but the standardised table structure provides a critical advantage for model training.