ValerianFourel/seoul-medical-facilities

Seoul Medical Facilities Dataset

Dataset Description

This dataset contains comprehensive information about unique medical facilities (hospitals, clinics) across all administrative districts (구) and neighborhoods (동) in Seoul, South Korea.

Note: This dataset contains only unique facilities. Duplicates have been removed based on place_id, with the most complete record retained for each facility.

Dataset Summary

• —Unique Facilities: 8,484
• —Districts Covered: 25
• —Neighborhoods (Dong) Covered: 320
• —Collection Period: 2025-12-28 to 2026-01-03
• —Source: Naver Maps
• —Language: Korean
• —Deduplication: Yes (by place_id)

Data Collection

Data was collected by systematically scraping Naver Maps for medical facilities across Seoul's administrative divisions:

• —Keywords: 병원 (hospital), 의원 (clinic), 클리닉 (clinic)
• —Coverage: All 25 districts (구) and 424+ neighborhoods (동)
• —Method: Automated web scraping with Selenium
• —Deduplication: Facilities appearing in multiple keyword searches are deduplicated by place_id

Facility Types

The dataset includes three types of medical facilities:

1. 1.병원 (Byeongwon) - Hospitals
2. 2.의원 (Uiwon) - Clinics/Medical offices
3. 3.클리닉 (Keullinik) - Specialized clinics

Note: Each facility appears only once, even if it matched multiple search keywords.

Dataset Structure

Data Fields

Important: place_id is the unique identifier. Each place_id appears exactly once in the dataset.

Geographic Coverage

Seoul's 25 districts (구):

• —Gangnam-gu, Gangdong-gu, Gangbuk-gu, Gangseo-gu, Gwanak-gu, Gwangjin-gu, Guro-gu, Geumcheon-gu, Nowon-gu, Dobong-gu, Dongdaemun-gu, Dongjak-gu, Mapo-gu, Seodaemun-gu, Seocho-gu, Seongdong-gu, Seongbuk-gu, Songpa-gu, Yangcheon-gu, Yeongdeungpo-gu, Yongsan-gu, Eunpyeong-gu, Jongno-gu, Jung-gu, Jungnang-gu

Usage

Load with Pandas

import pandas as pd

# Load parquet file
df = pd.read_parquet("hf://datasets/ValerianFourel/seoul-medical-facilities/seoul_medical_facilities.parquet")

# Basic exploration
print(f"Total unique facilities: {len(df):,}")
print(f"Districts: {df['file_district'].nunique()}")

# Each place_id is unique
assert df['place_id'].is_unique

# Filter by district
gangnam = df[df['file_district'] == 'Gangnam-gu']
print(f"Gangnam facilities: {len(gangnam):,}")

# Filter by facility type
hospitals = df[df['file_keyword'] == '병원']
print(f"Hospitals: {len(hospitals):,}")

Load with Datasets

from datasets import load_dataset

dataset = load_dataset("ValerianFourel/seoul-medical-facilities")
df = dataset['train'].to_pandas()

# Verify uniqueness
print(f"Unique facilities: {len(df):,}")
print(f"Unique place_ids: {df['place_id'].nunique():,}")
assert len(df) == df['place_id'].nunique()

Use Cases

• —Healthcare Access Analysis: Study distribution of medical facilities across Seoul
• —Geographic Analysis: Map healthcare infrastructure by district/neighborhood
• —Urban Planning: Identify underserved areas
• —Public Health Research: Analyze healthcare availability patterns
• —Business Intelligence: Market analysis for medical services
• —Navigation/Directory Apps: Build medical facility finders

Data Quality Notes

• —Deduplication: Each facility appears exactly once based on place_id
• —Completeness: For duplicate entries, the record with most complete information was retained
• —Unique Identifier: Use place_id to reference specific facilities
• —Phone numbers and websites may not be available for all facilities
• —Business hours may change; check official sources for current information
• —Review data is a snapshot at collection time

Limitations

• —Data represents a snapshot at collection time
• —Some fields may be incomplete (N/A values)
• —Limited to facilities discoverable via Naver Maps
• —Does not include detailed medical specialties or services
• —Operating hours and contact information may change

DATASET CONTEXT: Seoul Medical Facilities Knowledge Base

Regarding, facilitiesmetareviewsrag_ready.parquet: You have access to a structured database of medical facilities in Seoul, South Korea. Each record contains three types of data:

1. 1.Fact-Based Metadata: Name, address, hours, and parsed medical info.
2. 2.AI-Derived Accessibility Scores: Confidence scores (1-7) indicating if English services are available.
3. 3.Patient Sentiment Summaries: AI-generated summaries of thousands of patient reviews, grouped by topic.

COLUMN DEFINITIONS

• —Identity & Location
• —name: The official name of the facility.
• —category: Medical specialty (e.g., Dermatology, Dentistry).
• —address: Physical address in Seoul.
• —medical_info_parsed: (JSON/Dict) Specific procedures, equipment, or departments listed on their official profile.

• —Language Accessibility (AI-Scored)
• —english_confidence_score (1-7): The average confidence that staff speaks English, based on analyzing English/Mixed reviews.
• —1 = Definitely No, 4 = Ambiguous, 7 = Fluent/Verified.
• —english_max_score (1-7): The highest single piece of evidence found. High max + low average means "some staff might speak English, but it's not guaranteed."
• —has_english / has_mixed: (Boolean) True if actual reviews exist in these languages.

• —Patient Experience (RAG Data)
• —Summaries (List[str]): 3-10 sentence narrative summaries of patient feedback in English. Covers pros/cons, wait times, and kindness.
• —Key_Highlights (List[Dict]): Top semantic topics extracted from reviews (e.g., {'topic_en': 'Hidden Fees', 'relevance': 0.85}).
• —Summaries_Korean (List[str]): The same narrative summaries in Korean.

Citation

If you use this dataset, please cite:

@dataset{seoul_medical_facilities_2024,
  author = {Fourel, Valerian},
  title = {Seoul Medical Facilities Dataset},
  year = {2024},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/ValerianFourel/seoul-medical-facilities}
}

License

This dataset is released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license.

Maintenance

• —Maintainer: ValerianFourel
• —Last Updated: 2026-01-03

Acknowledgments

Data sourced from Naver Maps. This dataset is intended for research and educational purposes.