📌 1. Overview This dataset contains the structured transcription and semantic annotation of land registry records from a district of the commune of Gentilly (specifically the Butte-aux-Cailles neighborhood, which is today part of Paris). The data is extracted from three successive taxpayer mutation matrices (1813, 1836, and 1848), which correspond to two consecutive land registries versions used in this area, each featuring its own distinct set of maps and land plots numbering systems. This dataset was utilized to evaluate the PeGazUs algorithm, as described in a paper presented at the EKAW 2024 conference. The semantic annotation properties and elements are formally documented in the TABULAE ontology (as well as in the PeGazUs ontology). 📂 2. Repository contents (files) The dataset consists of 4 files in standard Comma-Separated Values (.csv) format, encoded in UTF-8: LISTE_PARCELLES.csv: An index file listing all land plots identified on the historical plots maps of the study area. It documents their presence or absence within the manually transcribed matrix records. Missing plots correspond to cases where the plot could not be found in the register. MAT_1813.csv: Transcription and semantic annotation of the 1813 cadastral matrix. MAT_1836.csv: Transcription and semantic annotation of the 1836 cadastral matrix. MAT_1848.csv: Transcription and semantic annotation of the 1848 cadastral matrix. 📊 3. Data structure and fields description The suffixes _transcript, _clean, and _treated denote different processing stages: raw transcription, normalized transcription, and semantic annotation respectively. 3.1. LISTE_PARCELLES.csv Field Name Description PLAN_PARCELLAIRE Publication year or reference of the historical cadastral map (e.g., 1810, 1847). Section Letter indicating the cadastral section (e.g., B, D). Parcelle The sequential plot number within its section. Bâti ? Built status of the plot (Oui/Non => True/False), indicating the presence of a building structure on the map. Trouvé-1813 / Trouvé-1836 / Trouvé-1848 Boolean flag (X if found, blank otherwise) certifying that at least one corresponding matrix line was successfully identified during the manual archival research phase. 3.2. MAT_1813, MAT_1836, MAT_1848.csv These files share a uniform, standardized schema designed to model the life cycle of land properties over time. Field Name Description Identifiers ID Internal identifier matching the reading sequence in the matrix during the manual transcription phase. UUID Globally unique identifier assigned to the plot mention during data processing and used in the generated Knowledge Graph. Register data Num_Folio Page or folio number in the original physical archive ledger. Type_CF Typology of the tax ledger account: Built or Non-Built property (Bâti / Non Bâti). Alt_Num_Folio Alternative folio number. Groupe_CF The land account index/number inside that specific folio. Ordre_de_lecture Sequential appearance order of the line on the original page. Image Archive reference id for the digitized document image from which the data is derived. Cote liée Reference shelfmark of a related image or document providing complementary context. Voie / Num_Voie When a folio indexes plots by address (rather than just by taxpayer), this indicates the thoroughfare name and street number if available. Plot data Section_clean Cleaned and standardized cadastral section identifier. Parcelle_clean / _treated The plot number (stripped of its section letter) in raw and normalized formats. Lieu-dit_transcript / _clean / _treated The land locality or address description, transitioning from abbreviated text to a standardized Street;Number format. Contribuable_transcript Textual transcription of the taxpayer/property owner description. Nature_transcript / _clean / _treated Land-use and property typology (House, Garden, Arable land, Washhouse, Shop, Workshop, etc.). Date entrée_transcript / _treated The transfer date into the specific tax account (raw transcription and standardized date format). Date sortie_transcript / _treated The transfer date out of the specific tax account (raw transcription and standardized date format). Tiré de_transcript / _treated Cross-reference pointing to the previous folio or plot state. Porté à_transcript / _treated Cross-reference pointing to the next folio or plot state. Ligne_barrée Boolean indicator (Oui/Non) showing if the entire transaction line is crossed out in the register. CF_rayé Boolean indicator (Oui/Non) showing if the entire tax account is crossed out in the register. Note on special transcription tokens: →: Indicates text wrapping/line breaks in the original manuscript. ~~text~~: Represents words or clauses that were crossed out/strikethrough in the physical book (corresponding to ×TEXT± in the information extraction dataset). ↑ and ↓: Denote respectively the start and end of superscript text (frequently utilized for historical abbreviations, e.g., Mme).