Matching Algorithm to Standardize Place Names in Casualty Lists of World War I

The German casualty lists from the First World War mention around 1.28 million different place names – far more places than actually existed at the time. Typos, misheard names and inconsistent spellings had turned single villages into dozens of variants. CompGen, Germany's largest genealogy association, had already digitised the lists, but couldn't reliably match the place names to the their Historic Geo Information System (GOV), which holds the correct historical names. Over four months in 2021, a team of five CorrelAid volunteers built compgen2, an open-source Python package that cleans up the entries, expands common abbreviations and uses phonetic and similarity matching to suggest the most likely GOV entry for each name. The team handed over the code and documentation to CompGen, including a catalogue of gaps and inconsistencies they had found in the GOV itself – useful input for the association's own data maintenance.