Bert2bert Base Arxiv Titlegen

🚀 Paper Title Generator

This project generates titles for computer science papers based on their abstracts, offering a useful tool for academic writing.

🚀 Quick Start

The model is a BERT2BERT Encoder - Decoder. It initializes the encoder and decoder using the official bert-base-uncased checkpoint. It was fine - tuned on 318,500 computer science papers posted on arXiv.org from 2007 to 2022. On the held - out validation data, it achieved a 26.3% Rouge2 F1 - Score.

✨ Features

• Accurate Title Generation: Given an abstract of a computer science paper, it can generate appropriate titles.
• High - Performance Model: Based on BERT2BERT Encoder - Decoder architecture and fine - tuned on a large number of papers, it shows good performance on validation data.

📚 Documentation

Model Details

The model is a BERT2BERT Encoder - Decoder. The encoder and decoder are initialized with the official bert-base-uncased checkpoint.

Training Data

It was fine - tuned on 318,500 computer science papers posted on arXiv.org between 2007 and 2022.

Evaluation Metrics

The model achieved a 26.3% Rouge2 F1 - Score on held - out validation data.

Live Demo

You can try the live demo at https://paper-titles.ey.r.appspot.com/.

📄 License

This project is licensed under the Apache 2.0 license.

Information Table