edsnlp
Extracts info from French clinical notes using rule-based and deep learning components.
Modular, fast NLP framework compatible with PyTorch and spaCy, built for French clinical texts.
EDS-NLP is an open-source framework designed to extract structured information from French clinical text. It combines spaCy-compatible rule-based pipes with PyTorch deep learning modules for tasks like negation detection and named entity recognition. It suits clinical researchers and developers working with French medical records.
What it is
The library organizes rule-based functions and deep learning modules into an efficient pipeline system. It supports multi-task deep learning models with weight sharing, LLM-based components, and fast multi-GPU inference. It also handles various data formats including BRAT, JSON, Parquet, Pandas, and Spark.
Who it's for
It suits developers and researchers building clinical data warehouse pipelines for French-language medical documents. It is not designed for general English-language clinical processing or users without Python programming experience.
Key features
French Clinical Rules
Rule-based components specifically tailored for French clinical notes.
PyTorch Integration
Compatible with PyTorch for trainable components like NER and span classification.
Multi-Task Weight Sharing
Supports multitask deep learning models with subcomponent weight sharing.
Flexible I/O Formats
Supports data loading and saving for BRAT, JSON, Parquet, Pandas, and Spark.
Up and running in 3 steps
Our honest verdict
What's great
- Native support for French clinical text extraction patterns
- Hybrid architecture combining spaCy rules with PyTorch deep learning
- Out-of-the-box multi-GPU support for fast inference
- Compatible with standard formats like JSON, Parquet, and Spark
Where it falls short
- Requires Python development skills to configure and run pipelines
- Performance depends heavily on the specific clinical population and documents
- Training machine learning components requires setting up PyTorch dependencies
Pricing & license
Free under its BSD-3-Clause license. The only cost is running it yourself on your own hardware infrastructure.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
Use it, or have us set it up
FAQ
What NLP backends does EDS-NLP use?
It uses spaCy for document representation and annotations, and PyTorch as the deep learning backend for trainable components.
Can I use it for languages other than French?
The framework is versatile and can be used on any textual document, but its pre-built rule components target French clinical notes.
