AgentPublic/local-administrations-directory

---

📢 Sondage 2026 : Utilisation des datasets publiques de MediaTech

Vous utilisez ce dataset ou d’autres datasets de notre collection MediaTech ? Votre avis compte ! Aidez-nous à améliorer nos datasets publiques en répondant à ce sondage rapide (5 min) : 👉 https://grist.numerique.gouv.fr/o/albert/forms/gF4hLaq9VvUog6c5aVDuMw/11 Merci pour votre contribution ! 🙌

---

🇫🇷 French Local Administrations Directory Dataset

This dataset is a processed and embedded version of the public data Annuaire de l’administration - Base de données locales (French Local Administrations Directory), published on data.gouv.fr.
This information is also available on the official directory website of Service-Public.fr: https://lannuaire.service-public.fr/

The dataset provides semantic-ready, structured and chunked data of French local public entities, including organizational details, missions, contact information, and hierarchical links. Each chunk of text is vectorized using the BAAI/bge-m3 embedding model to enable semantic search and retrieval tasks.

---

🗂️ Dataset Contents

The dataset is provided in Parquet format and contains the following columns:

Column Name	Type	Description
chunk_id	str	Unique source based identifier of the chunk
doc_id	str	Document identifier. Identical to chunk_id as each document only has 1 chunk.
chunk_xxh64	str	XXH64 hash of the chunk_text value.
types	str	Type(s) of administrative entity.
name	str	Name of the organization or service.
mission_description	str	Description of the entity's mission.
addresses	list[dict]	List of address objects (street, postal code, city, etc.).
phone_numbers	list[str]	List of telephone numbers.
mails	list[str]	List of contact email addresses.
urls	list[str]	List of related URLs.
social_medias	list[str]	Social media accounts.
mobile_applications	list[str]	Related mobile applications.
opening_hours	str	Opening hours.
contact_forms	list[str]	Contact form URLs.
additional_information	str	Additional information.
modification_date	str	Last update date.
siret	str	SIRET number.
siren	str	SIREN number.
people_in_charge	list[dict]	List of responsible persons.
organizational_chart	list[str]	Organization chart references.
hierarchy	list[dict]	Links to parent or child entities.
directory_url	str	Source URL from the official state directory website.
chunk_text	str	Textual content of the administrative chunk.
embeddings_bge-m3	str (stringified list)	Embeddings of chunk_text using BAAI/bge-m3. Stored as a JSON array string.

---

🛠️ Data Processing Methodology

📥 1. Field Extraction

The following fields were extracted and/or transformed from the original JSON:

• Basic fields: chunk_id, doc_id, name, types, mission_description, additional_information, siret, siren, directory_url, modification_date are directly extracted from JSON attributes.
• Structured lists:
  • addresses: list of dictionaries with adresse, code_postal, commune, pays, longitude, and latitude.
  • phone_numbers, mails, urls, social_medias, mobile_applications, contact_forms: derived from their respective fields with formatting.
• People and structure:
  • people_in_charge: list of dictionaries representing staff members or leadership (title, name, rank, etc.).
  • organizational_chart, hierarchy: structural information within the administration.
• Other fields:
  • opening_hours: built using a custom function that parses declared time slots into readable strings.
  • chunk_xxh64: is the xxh64 hash of the chunk_text value. It is useful to determine if the chunk_text value has changed from a version to another.

✂️ 2. Generation of chunk_text

A synthetic text field called chunk_text was created to summarize key aspects of each administrative body. This field is designed for semantic search and embedding generation. It includes:

• The entity’s name : name
• Its mission statement (if available) : mission_description
• Key responsible individuals (formatted using role, title, name, and rank) : people_in_charge

There was no need here to split characters here.

🧠 3. Embeddings Generation

Each chunk_text was embedded using the BAAI/bge-m3 model. The resulting embedding vector is stored in the embeddings_bge-m3 column as a string, but can easily be parsed back into a list[float] or NumPy array.

🎓 Tutorials

🤖 1. How to load MediaTech's datasets from Hugging Face and use them in a RAG pipeline ?

To learn how to load MediaTech's datasets from Hugging Face and integrate them into a Retrieval-Augmented Generation (RAG) pipeline, check out our step-by-step RAG tutorial available on our GitHub repository !

📌 2. Embedding Use Notice

⚠️ The embeddings_bge-m3 column is stored as a stringified list of floats (e.g., "[-0.03062629,-0.017049594,...]"). To use it as a vector, you need to parse it into a list of floats or NumPy array.

Using the datasets library:

import pandas as pd
import json
from datasets import load_dataset
# The Pyarrow library must be installed in your Python environment for this example. By doing => pip install pyarrow

dataset = load_dataset("AgentPublic/local-administrations-directory")
df = pd.DataFrame(dataset['train'])
df["embeddings_bge-m3"] = df["embeddings_bge-m3"].apply(json.loads)

Using downloaded local Parquet files:

import pandas as pd
import json
# The Pyarrow library must be installed in your Python environment for this example. By doing => pip install pyarrow

df = pd.read_parquet(path="local-administrations-directory-latest/") # Assuming that all parquet files are located into this folder
df["embeddings_bge-m3"] = df["embeddings_bge-m3"].apply(json.loads)

You can then use the dataframe as you wish, such as by inserting the data from the dataframe into the vector database of your choice.

📚 Source & License

🐱 GitHub repository :

The project MediaTech is open source ! You are free to contribute or see the complete code used to build the dataset by checking the GitHub repository

🔗 Source :

• Lannuaire.Service-Public.fr
• Data.Gouv.fr : Service-public.fr - Annuaire de l’administration - Base de données locales

📄 Licence :

Open License (Etalab) — This dataset is publicly available and can be reused under the conditions of the Etalab open license.