Research Paper Classification and Publishability Prediction

This project classifies research papers based on their content and predicts their publishability. The model determines whether a paper is suitable for publication and, if so, suggests the most relevant conference.

Features

• Uses SciBERT for embedding paper content
• Self-training classifier for publishability prediction
• Sentence-BERT for similarity-based conference classification
• Supports multiple conferences, including CVPR, NeurIPS, EMNLP, TMLR, and KDD

Prerequisites

Ensure you have the following installed:

• Python 3.8+
• Required dependencies (install via requirements.txt)

Installation

1. Clone the repository:

   git clone https://github.com/Ayush-Sharma23/pathway-hackathon-resources.git
   cd pathway-hackathon-resources

2. Install dependencies:

   pip install -r requirements.txt

Usage

1. Prepare CSV files:
   • data/labeled_data.csv (contains labeled paper content and labels for training)
   • data/unlabeled_data.csv (contains paper content for classification)
2. Run the script:

   python main.py

3. The results will be saved in results/output.csv.

File Structure

research-paper-classifier/
│── main.py          # Main script for training and classification
│── requirements.txt # Dependencies
│── data/            # Folder containing input data
│   ├── labeled_data.csv
│   ├── unlabeled_data.csv
│── results/         # Folder to store outputs
│   ├── output.csv
│── README.md        # Project documentation

Dependencies

• pandas
• numpy
• transformers
• sentence-transformers
• scikit-learn
• tqdm

Future Improvements

• Fine-tune SciBERT for better embeddings
• Add more conferences and their topic embeddings
• Improve classification accuracy using advanced ML techniques