PErpetual GAZeteer of approach-address UtteranceS

This repository contains the documentation of the PeGaZus ontology and knowledge graph construction method.

Abstract

Gazetteers, as compilation of named places, are central resources on the Web of data, as they provide a common ground to link and integrate many textual or structured resources on the Web. Gazetteers usually categorise and associate places names with geospatial coordinates. In more recent times, historical gazetteers, which aim to represent places from the past, have received increasing attention. The creation of these gazetteers poses specific challenges, including the definition of the identity of evolving places, the representation of their evolution through time (how they change, when the changes happen), and the population of the gazetteer based on scarce and heterogeneous historical sources.

We propose an approach to create an urban historical gazetteer on the evolution of two major urban large-scale types of places, namely addresses and land plots. Our proposal is inspired by approaches for creating knowledge graphs and takes advantage of the knowledge representation and reasoning possibilities offered by Semantic Web standards to address the aforementioned challenges. The approach was applied to the \textit{Butte aux Cailles} district of Paris, for which a variety of contemporary and historical sources were used. The resulting knowledge graph can be used for a variety of purposes, including historical geocoding of old documents, identifying the use of a plot of land at a given date, and recording the events that led to its current state.

Structure of the repository

├── data                      <- RDF resources used to build the graph
│   ├── addresses                   <- raw resources for the specific case of addresses and streets
│   ├── land-registry               <- RDF resources for the specific case of the plots from the Napoleonic land registry
│
├── ontology                  <- PeGaZus Ontology
│   ├── ontology-addresses.ttl       <- Core part of the ontology to describe landmarks and addresses
│   ├── ontology-land-registry.ttl   <- Specific modules used to describe the Napoleonic land registry documents and landmarks
│   ├── documentation
│       ├── addresses
│           ├── land_registry_landmarks_extension
│       ├── land_registry_documents_use
│       ├── sources
│       ├── taxpayers
│       ├── temporal_evolution
│
├── scripts
│   ├── addresses                    <- Implementation of the algorithm specialized for the addresses
│   ├── land-registry                <- Implementation of the algorithm specialized for the Napoleonic land registry
│
├── LICENCE.md
└── README.md

data folder

This folder stores files used to build knowledge graph. It is split in two repositories:

• addresses folder contains csv and geojson files which describe addresses and streets from different sources at different times (RDF resources are built during process) ;
• land-registry has RDF resources for the specific case of the plots from the Napoleonic land registry.

⚠️ To get more information about their content, please read their readme : addresses and land-registry

ontology folder

Two files describe the ontology: ontology-addresses.ttl is the core part of the ontology and describe landmarks and addresses whereas ontology-land-registry.ttl integrates specific modules used to describe the Napoleonic land registry documents and landmarks.

Ontology documentation is divided into as many parts as there are modelets:

• addresses with its extension for land registry landmarks ;
• land_registry_documents_use ;
• sources ;
• taxpayers ;
• temporal_evolution.

Each modelet documentation has 3 or 4 files:

• {modelet_name}_scenario.md: the natural-language argument describing the sub-problem to be addressed ;
• {modelet_name}_glossary.md: glossary which defines the main terms involved ;
• {modelet_name}_competency_questions.md: set of informal competence questions (in natural language) which represent the questions to be answered by the knowledge base ;
• {modelet_name}_sparql_queries.md: it translates informal competence questions into SPARQL queries.

scripts folder

This folder contains code to build knowledge graphs. Since addresses and land plots have their own specificities, each one has a folder in which its code allow KG construction.

⚠️ To get more information about their content, please read their readme : addresses and land-registry

Acknowledgements

This work was supported by the IGN and the AID.

To cite this work

@inproceedings{bernard:hal-04721538,
  TITLE = {{PeGazUs: A knowledge graph based approach to build urban perpetual gazetteers}},
  AUTHOR = {Bernard, Charly and Tual, Solenn and Abadie, Nathalie and Dum{\'e}nieu, Bertrand and Perret, Julien and Chazalon, Joseph},
  URL = {https://hal.science/hal-04721538},
  BOOKTITLE = {{International Conference on Knowledge Engineering and Knowledge Management (EKAW 2024)}},
  ADDRESS = {Amsterdam, Netherlands},
  PUBLISHER = {{Springer Nature Switzerland}},
  SERIES = {Lecture Notes in Computer Science},
  VOLUME = {15370},
  PAGES = {364-381},
  YEAR = {2024},
  MONTH = Nov,
  DOI = {10.1007/978-3-031-77792-9\_22},
}