NOL-Event-Data-Classifier

A machine learning framework designed to format, classify, and fine-tune event labels retrieved from NOL (Nociception Level) monitoring devices.

Features

• Label Classification: Combines supervised and unsupervised learning techniques to categorize data based on format.
• Intraoperative Focus: Provides specialized processing for intraoperative anesthesia workflows (labels.py).
• Model Fine-Tuning: Includes dedicated scripts to fine-tune trained models for downstream tasks.

Integration

This project serves as the default supervised classification engine for ORRT (OR-Recorder-Transcriber)—a desktop application for real-time audio recording and AI transcription—specifically for intraoperative anesthesia data.

---

Run the project

[!WARNING] Requires Python 3.12 or higher (tested on 3.12.13 which is highly recommanded)

Installation

Get the project

git clone https://github.com/Gabrieleirbag1/NOL-Event-Data-Classifier.git

cd NOL-Event-Data-Classifier

Install dependancies

pip install -r requirements.txt

Run Python scripts

# Supervised
python3 src/nol_event_classifier/supervised/main.py

# Unsupervised
python3 src/nol_event_classifier/unsupervised/main.py

# Fine Tuning (Supervised)
# -m: Base model name for SetFit
# -o: Output name
# -t: Test size for train/test split
# -e: Number of epochs
# -b: Batch size
# -i: Number of iterations
python3 src/nol_event_classifier/supervised/fine_tuning.py \
  -m paraphrase-multilingual-MiniLM-L12-v2 \
  -o "setfit_paraphrase-multilingual-mpnet-base-v2_3ep-16bs-10it" \
  -t 0.2 \
  -e 3 \
  -b 16 \
  -i 10

Author

@Missclick (Developer) E-mail : [email protected] Discord : missclick.net