---
title: Clinical NER Pipeline Comparison
emoji: 🧠
colorFrom: blue
colorTo: green
sdk: docker
app_port: 7860
pinned: false
license: gpl-3.0
short_description: Comparison of strategies for NER.
---

# Clinical NER Pipeline Comparison

This demo compares three approaches to clinical entity recognition:

1. Fine-tuned clinical BERT (NER)
2. Vanilla BERT embeddings + similarity
3. Static Word2Vec embeddings + similarity

The goal is to demonstrate why **fine-tuning and context matter**.

## How to use
- Select a predefined sentence or type your own
- Adjust prototype words if desired
- Click **Execute**
- Compare the outputs of the three pipelines

## Notes
- Word2Vec is large and may take time to load on first run
- This is a didactic comparison, not a production NER system