Add clinical summarisation tutorial using synthetic discharge notes

[Rifa-111]

Overview

This PR adds a new tutorial chapter demonstrating clinical text summarisation using synthetic discharge notes and a T5-style transformer model.

The tutorial covers:

• generating synthetic discharge summaries (no PHI or real patient data)
• preparing the data for sequence-to-sequence summarisation
• fine-tuning google/flan-t5-small on the synthetic dataset
• evaluating the model with ROUGE using the evaluate library
• discussing limitations, ethical considerations and extensions

Motivation

Clinical summarisation is a common real-world use case for transformers, but publicly available datasets containing real discharge notes are restricted due to privacy concerns. This tutorial shows how to prototype workflows safely using synthetic data while still teaching the core modelling, evaluation and reproducibility concepts.

Implementation details

• New chapter: chapters/clinical_summarisation.md
• Uses:
  • datasets to handle tabular text data
  • transformers for model loading and fine-tuning
  • evaluate for summarisation metrics (ROUGE)
• Designed to be runnable on CPU or a single GPU / Colab environment.

Notes

• All examples use synthetic discharge notes generated from templates; no PHI or real patient-identifiable information is used.
• Happy to adjust naming, placement or style to better align with the course structure and Hugging Face documentation conventions.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.