CoolFace
Datasetpublic

opendoc-pakistan/pakistan_drug_registry

Pakistan Pharmaceutical Product Registry (RAG Corpus) A retrieval-augmented-generation (RAG) ready corpus of medicines relevant to Pakistan. Two parts: (1) marketed pharmaceutical products with retail prices, pack sizes and availability; (2) Pakistan's National Essential Medicines List (WHO Model List, 23rd List 2023) with therapeutic category, listing type and dosage forms/strengths. Each part ships as JSONL (embeddable passage + structured metadata) plus a flat CSV twin.… See the full description on the dataset page: https://huggingface.co/datasets/opendoc-pakistan/pakistan_drug_registry.

sourceHugging Facemitupdated 3mo agoView on Hugging Face
0likes41downloads
Dataset Card

Pakistan Pharmaceutical Product Registry (RAG Corpus)

A retrieval-augmented-generation (RAG) ready corpus of medicines relevant to Pakistan. Two parts: (1) marketed pharmaceutical products with retail prices, pack sizes and availability; (2) Pakistan's National Essential Medicines List (WHO Model List, 23rd List 2023) with therapeutic category, listing type and dosage forms/strengths. Each part ships as JSONL (embeddable passage + structured metadata) plus a flat CSV twin.


Source: OpenDoc Pakistan

Pakistan Pharmaceutical Product Registry (RAG Corpus)

Summary

A retrieval-augmented-generation (RAG) ready corpus of medicines relevant to Pakistan, in two parts:

  1. 1.Marketed products — pharmaceutical products sold in Pakistan with retail prices (before/after discount), pack sizes, manufacturers, and market availability (321 products).
  2. 2.Essential medicines — Pakistan's National Essential Medicines List (WHO Model List, 23rd List 2023), with therapeutic category, listing type (core/complementary), and dosage forms/strengths (398 medicines).

Each part ships in two equivalent formats: a JSONL corpus (one embeddable passage plus structured metadata per record) for RAG / vector-search pipelines, and a flat CSV for tabular ML and quick validation. This is the reference dataset defining OpenDoc's RAG corpus convention.

Dataset Details

  • —Dataset ID: pakistan-drug-registry
  • —Version: 1.0.0
  • —Category: Health
  • —Tags: drugs, medicines, pharmaceutical, prices, rag, healthcare, pakistan
  • —Language: English
  • —Geography: Pakistan
  • —File Formats: JSONL, CSV
  • —Records: 321 marketed products (83 companies) + 398 essential medicines (19 therapeutic categories)
  • —Version: 1.1.0
  • —Update Frequency: Irregular (as sources are refreshed)
  • —Maintainer: OpenDoc Data Stewardship Team

Source and Provenance

  • —Source Name: Open Data Pakistan (ODP) — "Pharmaceutical product pricing and availability in Pakistan" (originally published on Kaggle by Talha Sattar)
  • —Source URL: https://opendata.com.pk/dataset/pharmaceutical-product-pricing-and-availability-in-pakistan
  • —Upstream Source: https://www.kaggle.com/datasets/talhasattar727/pakistan-pharmaceutical-dataset
  • —Source Type: Open-data portal CSV resource (CKAN), sourced from a Kaggle dataset (secondary research)
  • —Collection Method: Direct download of the published CSV resource.
  • —Collection Date: July 2026
  • —Processing Pipeline: Deduplicated rows, parsed prices into integer PKR amounts, parsed discount percentages, generated a stable id per product, and synthesised a natural-language text passage per record for embedding. Emitted both JSONL and CSV.
  • —Raw Data Availability: Public on the ODP portal.

License

  • —Marketed products — Original License: MIT (as declared on the upstream Kaggle source). Redistributed under MIT.
  • —Essential medicines — Source: Government of Pakistan gazette notification S.R.O. 1423(I)/2023 (public statutory notification) adopting the WHO Model List of Essential Medicines. WHO Model List content is © WHO, reproduced here for reference from the public gazette; WHO material is generally reusable with attribution under WHO's terms.
  • —Attribution Required: Yes
  • —Commercial Use: Allowed (marketed-products part under MIT)
  • —Redistribution: Allowed
  • —Restrictions: Retain the MIT notice and credit Talha Sattar & Open Data Pakistan (products); credit MoNHSR&C / WHO (essential medicines); and credit OpenDoc Pakistan.

Intended Use

Appropriate for:

  • —Building RAG chatbots / assistants that answer questions about Pakistani medicine prices and availability.
  • —Vector-search and semantic-retrieval demos over a domain corpus.
  • —Tabular price analysis and ML (via the CSV twin).

Out-of-Scope Use

  • —Not medical advice. Contains no dosing, indication, contraindication, or safety information.
  • —Not a complete or official drug registry — it is a market price/availability snapshot, not the DRAP master registration list.
  • —Prices are point-in-time and will drift; do not use for current billing.

Data Structure

Marketed products (source: Open Data Pakistan / Kaggle, MIT):

  • —data/processed/drug_registry.jsonl — one JSON object per line: {id, text, metadata{product_name, company, pack_size, price_before_pkr, discount_percent, price_after_pkr, availability, country, source}}.
  • —data/processed/drug_registry.csv — flat equivalent with the metadata fields plus the text column.

Essential medicines (source: Pakistan gazette S.R.O. 1423(I)/2023 adopting the WHO Model List of Essential Medicines, 23rd List 2023):

  • —data/processed/essential_medicines.jsonl — {id, text, metadata{medicine, therapeutic_category, subcategory, list_type, dosage_forms, country, source}}.
  • —data/processed/essential_medicines.csv — flat equivalent.

The essential-medicines part was extracted from the WHO-hosted PDF of the Pakistan NEML gazette using automated PDF parsing (firecrawl parse), then structured by therapeutic section.

Data Dictionary

See data_dictionary.csv.

Quality Report

  • —Quality Score: 99 / 100
  • —Rating: Excellent
  • —Records: 321 unique products (exact and price/pack duplicates removed).
  • —Duplicate Rows: 0% (primary key: id).
  • —Validation Status: CSV twin passes Frictionless tabular-data-resource schema.
  • —Known Issues: Some products share a brand name with different pack sizes/prices (kept as distinct records). Company names are as-listed and not normalised to legal entity names.

Privacy and Sensitivity Review

  • —Privacy Level: P0 (commercial product listings; no personal data)
  • —Personal Data Present: No
  • —Sensitive Data Present: No
  • —Redactions Applied: None

Bias, Gaps, and Limitations

  • —Marketed-products coverage is limited to the source snapshot; not exhaustive of the Pakistani market.
  • —The marketed-products and essential-medicines parts are not cross-linked (brand↔generic mapping is out of scope for this version).
  • —Essential-medicines entries were extracted via automated PDF parsing; dosage-form strings are as-listed and may contain minor OCR/formatting artefacts.

Example Usage

python
import json

# Load the RAG corpus
with open("data/processed/drug_registry.jsonl", encoding="utf-8") as f:
    docs = [json.loads(line) for line in f]

# Each doc is ready to embed
print(docs[0]["text"])
print(docs[0]["metadata"]["price_after_pkr"])

Citation

text
Open Data Pakistan, "Pharmaceutical product pricing and availability in Pakistan", curated as "Pakistan Pharmaceutical Product Registry (RAG Corpus)" by OpenDoc Pakistan (OpenDoc), v1.0.0, 2026.