Multi-Document Summarization with Centroid-Based Pretraining

This project implements the methodology described in the paper Multi-Document Summarization with Centroid-Based Pretraining. It focuses on improving multi-document summarization by pretraining models using centroid-based objectives and fine-tuning them for summarization tasks.

🚀 Features

• Centroid-Based Pretraining: Generates pseudo-label summaries from document clusters.
• Fine-Tuning: Adapts pretrained models to generate human-like summaries.
• Evaluation: Measures performance using ROUGE and other metrics.
• Modular code for reproducibility and experimentation.

🛠️ Installation

1. Clone the repository: git clone https://github.com/username/Multi-Document-Summarization.git cd Multi-Document-Summarization 02. Install dependencies: pip install -r requirements.txt 03. Download and prepare datasets:

• Place raw datasets in data/raw/.
• Follow instructions in data/README.md for preprocessing.