AshenFdo/synthetic_blood_request_urgency_dataset

Synthetic Blood Request Urgency Dataset

A synthetic dataset of Sri Lanka-based blood donation requests, labeled for urgency classification (emergency / not_emergency). Built to train a text classification model that powers a real-world blood donation application targeting Sri Lanka.

🎯 Purpose

Blood donation platforms receive a mix of requests — some are life-threatening emergencies requiring immediate action, and others are routine or planned needs. Treating every request the same wastes critical response time. This dataset was created to train a binary text classification model that can automatically detect whether a blood request is an emergency or not an emergency, based on its description. The model is intended to be integrated into a blood donation mobile application for Sri Lanka, where it will help: - Prioritize urgent requests and surface them to donors first - Send targeted alerts to nearby donors for life-critical cases - Reduce response time for patients in critical condition

🌍 Why Sri Lanka Only?

The dataset is intentionally scoped to Sri Lanka because the target application serves Sri Lankan donors and patients. All descriptions reference Sri Lankan hospitals (e.g., Karapitiya, Kurunegala, Kandy National, Ratnapura), cities, and locally relevant blood collection contexts. This geographic focus ensures the model learns language patterns specific to how Sri Lankan users describe blood needs.

📊 Dataset Summary

Property	Value
Total Samples	2,500
Classes	emergency, not_emergency
Class Balance	50% / 50% (1,250 each)
Language	English
Data Type	Synthetic (AI-generated)
Domain	Healthcare / Blood Donation
Geography	Sri Lanka

---

🏷️ Label Definitions

Label	Description
emergency	Requests where a patient is in a critical or life-threatening condition requiring immediate blood. Typically includes language like "urgent", "critical", "out of time", "severe hemorrhage", or hospital emergency notices.
not_emergency	Routine, planned, or replacement donation requests. Includes scheduled surgeries, chronic condition support, community drives, and replacement donor programs.

---

💡 Example Rows

Description	Label
"HOSPITAL EMERGENCY: Ratnapura Hospital urgently requires AB- blood for a critical patient."	emergency
"O negative blood required urgently at Karapitiya Hospital. We are out of time."	emergency
"Organizing AB+ blood for an upcoming planned operation at Kandy National Hospital."	not_emergency
"Our Lions Club is fulfilling a replacement donor program for a patient at Anuradhapura."	not_emergency
"Looking for regular B- donors to support a patient with a chronic blood disorder at Jaffna Teaching Hospital."	not_emergency

---

📁 Data Fields

Field	Type	Description
description	string	Free-text blood request description
label	string	Urgency class: emergency or not_emergency

---

🚀 Intended Use

Primary Use

Training and evaluating a text classification model for blood request urgency detection, to be deployed in a Sri Lanka blood donation application. ### Planned HuggingFace Space A simple demo Space will accompany this dataset, where you can paste a blood request description and see whether the model classifies it as an emergency or not. ### Other Potential Uses - Benchmarking small NLP models on healthcare urgency detection - Fine-tuning lightweight models (e.g., DistilBERT, MobileBERT) for edge deployment - Studying how urgency language patterns differ in South Asian healthcare contexts

⚠️ Limitations & Biases

• Synthetic data — All samples were AI-generated. Real-world requests may use different phrasing, abbreviations, Sinhala/Tamil mix, or informal language not captured here.
• English only — Sri Lankan blood requests often appear in Sinhala or Tamil. This dataset does not cover those languages.
• Binary labels only — Urgency in practice exists on a spectrum. A two-class label may not capture all nuance.
• Not for clinical use — This dataset and any model trained on it should not be used as a substitute for medical triage or clinical decision-making.

---

🔮 Roadmap

• Expand dataset with real anonymized requests (with user consent)
• Add Sinhala and Tamil language samples
• Train and publish a fine-tuned model to HuggingFace Hub
• Launch a demo Space for live inference
• Integrate model into Sri Lanka blood donation mobile app

---

📜 License

This dataset is released under the MIT License. You are free to use, modify, and distribute it with attribution.

🙏 Citation

If you use this dataset in your research or project, please cite it as: @dataset{AshenFdo_synthetic_blood_request_urgency_2025, author = {AshenFdo}, title = {Synthetic Blood Request Urgency Dataset}, year = {2025}, publisher = {HuggingFace}, url = {https://huggingface.co/datasets/AshenFdo/synthetic_bood_request_urgency_dataset} }

Built with the goal of saving lives — one donation at a time. 🩸