[Dataset] A dataset for named entity recognition on taxpayers mentions extracted from the 19th-century french land registry

This is a dataset to train/fine-tune Named Entity Recognition (NER) models to recognise taxpayers properties from taxpayers mentions in the tables of the 19th century French Land Registry. The dataset is provided in the Brat Standoff Format. NER Classes name : lastname of the taxpayer or organisation name firstnames : 1...n firstnames of the taxpayer activity : taxpayer activity address : taxpayer address title : taxpayer title (e.g. : prince) familystatus : taxpayer family status (like father, mother etc.) Special tokens Text includes some special tokens that are used to represent additionnal layout informations like : → : back to a new line ↑TEXT↓ : exponent ×TEXT± : crossed out text (most of the time, means outdated or erroneous information according the 19th-century administrators of the land registry) Important notes The "taxpayers entries" of this dataset correspond to all "not ditto/not empty" cells of the IE dataset available at the following address : 10.5281/zenodo.15411507 v0 is a "quickly annotated dataset" : pre-annotated with few-shot learning with llama3.1, manually corrected (but deeper curation still required) and contains several mentions of the same taxpayers in the different subsets (train, val, test). v0 is the version used in the following TPDL 2025 article : 10.1007/978-3-032-05409-8_24 To create the subsets, see the code in the Git-Hub repository bellow.