README.md

metadata

pretty_name: Miam Paris Original Data
language:
  - fr
  - en
tags:
  - paris
  - restaurants
  - establishments
  - geospatial

Miam_Paris_Original_Data

Full collection of Paris food & drink establishments (restaurants, cafés, bars, bakeries, tea shops...), organized by operating status and data-access category. Mirrors the Miam team's cleaned local Original Data folder (state of 2026-07-10).

Structure

open/
  Public Access/            # publicly accessible venues (anyone can eat there)
  Restricted Access/        # hotel-guest-only, corporate/staff, members-only, etc.
  Uncertain Restriction/    # unclear public-access status
temporarily_closed/         # temporarily closed establishments
uncertain/                  # uncertain operating status

Each establishment folder contains:

<Establishment_Name>/
  exploration_data.json   # name, address + coordinates, contact, hours, menu, ratings,
                          #   restricted_access classification, verified_status...
  photos/                 # facade, interior, food, menu photos (jpg/png/webp)
  menus/  documents/      # menu PDFs / documents when available

Statistics (2026-07-10)

Category	Venues
open / Public Access	8,769
open / Restricted Access	280
open / Uncertain Restriction	6
temporarily_closed	50
uncertain	59
Total	9,164

Data fields of note

• address.coordinates — latitude/longitude, geocoded/validated against the IGN/BAN geocoder (~100% coverage outside closed; a handful of genuinely-suburban venues sit outside the Paris bounding box by design).
• restricted_access — {status: no|yes|uncertain, confidence, reason_code (R1–R7), reason}, classifying whether the general public can eat there. no → Public Access, yes → Restricted Access, uncertain → Uncertain Restriction.
• ratings — weighted/estimated rating computed from multi-source reviews (where available).
• verified_status — operating-status verification (open / closed / temporarily_closed / uncertain) with confidence.

Cleaning notes

• Deduplicated on normalized establishment_name + address.full_address, then further by canonical geocoded coordinate (IGN/BAN): venues resolving to the same physical point with the same name (including word-order variants and asset-only crawl duplicates) were collapsed to one. Same-name establishments at genuinely different addresses (chain branches) are kept as separate folders. Distinct venues sharing one building (e.g. a hotel's restaurant + bar, a mall's outlets) are deliberately kept separate.
• Venues classified into the three open/ access sub-categories by an LLM public-accessibility triage over the restricted_access fields.
• Private dataset for the Miam team.