https://github.com/meetptl04/pretox-classifier

BioBERT-based text classifier to predict PRETOX-related text (PRETOX_REL vs NO_PRETOX_REL).
https://github.com/meetptl04/pretox-classifier

biobert biomedical deep-learning huggingface machine-learning nlp pytorch streamlit text-classification transformer-model

Last synced: about 2 months ago
JSON representation

BioBERT-based text classifier to predict PRETOX-related text (PRETOX_REL vs NO_PRETOX_REL).

• Host: GitHub
• URL: https://github.com/meetptl04/pretox-classifier
• Owner: meetptl04
• License: mit
• Created: 2025-08-16T12:55:25.000Z (10 months ago)
• Default Branch: main
• Last Pushed: 2025-08-31T18:37:18.000Z (9 months ago)
• Last Synced: 2025-08-31T20:33:43.549Z (9 months ago)
• Topics: biobert, biomedical, deep-learning, huggingface, machine-learning, nlp, pytorch, streamlit, text-classification, transformer-model
• Language: Jupyter Notebook
• Homepage:
• Size: 523 KB
• Stars: 0
• Watchers: 0
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# PRETOX Text Classifier with BioBERT

![Python](https://img.shields.io/badge/Python-3.9+-blue?style=flat-square&logo=python)

![Streamlit](https://img.shields.io/badge/Streamlit-App-red?style=flat-square&logo=streamlit)

![BioBERT](https://img.shields.io/badge/Model-BioBERT-green?style=flat-square&logo=semanticweb)

![Accuracy](https://img.shields.io/badge/Accuracy-94%25-brightgreen?style=flat-square)

![License](https://img.shields.io/badge/License-MIT-yellow?style=flat-square)

A high-performance biomedical text classifier built with **BioBERT** to determine if a given text is related to **PRETOX** (Preclinical Toxicology). This tool is designed to help researchers efficiently sift through vast amounts of literature to find toxicology-relevant information.

---

## Key Features

- **High Accuracy**: Achieves **94% accuracy** on the test dataset, ensuring reliable classification.

- **Domain-Specific AI**: Utilizes **BioBERT**, a language model pre-trained on biomedical text, for nuanced understanding of scientific terminology.

- **Interactive Interface**: A simple and intuitive web application built with Streamlit allows for easy, code-free use by researchers and domain experts.

- **Reproducible**: The entire training and evaluation pipeline is available in a Jupyter Notebook for full transparency and custom experimentation.

---

## Live Demo

Here is a preview of the interactive Streamlit application. Users can input text directly or use the provided examples to get an instant classification.

![Streamlit App Demo](images/streamlit_app_image.png)

---

## Model Performance

The model demonstrates strong performance in distinguishing between PRETOX-related and non-related texts.

**Final Test Accuracy**: **94%**

### Classification Report

| Class             | Precision | Recall | F1-Score |

| ----------------- | :-------: | :----: | :------: |

| `0 (NO_PRETOX_REL)` |   0.97    |  0.93  |   0.95   |

| `1 (PRETOX_REL)`  |   0.91    |  0.96  |   0.93   |

### Training & Validation Curves

The training history shows stable learning and good generalization from training to validation data.

![Accuracy and Loss Curves](images/accuracy.png)

---

## Getting Started

Follow these steps to set up and run the project locally.

### 1. Clone the Repository

```

git clone https://github.com/meetptl04/pretox-classifier.git

cd pretox-classifier

```

### 2\. Set Up a Virtual Environment

It is recommended to use a virtual environment to manage dependencies.

  - **Linux / macOS:**

    ```bash

    python3 -m venv venv

    source venv/bin/activate

    ```

  - **Windows:**

    ```bash

    python -m venv venv

    venv\Scripts\activate

    ```

### 3\. Install Dependencies

```bash

pip install -r requirements.txt

```

### 4\. Run the Streamlit App

```bash

streamlit run app.py

```

Navigate to the local URL provided in your terminal to start using the application.

-----

## Model & Dataset

  - **Model**: The classifier is a fine-tuned version of [**BioBERT (v1.1)**](https://huggingface.co/dmis-lab/biobert-v1.1), optimized for sequence classification. To review or retrain the model, please see the `NLP_BioBert_PRETOX_REL.ipynb` notebook.

  - **Dataset**: The model was trained on the [**pretoxtm-dataset**](https://huggingface.co/datasets/javicorvi/pretoxtm-dataset) from Hugging Face, which contains biomedical text excerpts labeled for toxicology relevance.

-----

## Project Structure

```

BioBert_app/

├── data/                    # Dataset files

├── model/                   # Trained model checkpoints

├── images/                  # Project images and screenshots

│   ├── accuracy.png

│   └── streamlit_app_image.png

├── venv/                    # Virtual environment (ignored by git)

├── app.py                   # Streamlit application source code

├── requirements.txt         # Required Python packages

└── NLP_BioBert_PRETOX_REL.ipynb # Notebook with training & evaluation code

```

-----