Dialogrpt Depth

🚀 Demo

This is a demonstration project that provides a practical example of using DialogRPT-depth for dialog response ranking. You can try out the 📔 Colab Notebook Demo (click me!).

💻 Usage Examples

Basic Usage

The following table shows some example dialogs and their corresponding depth scores:

The depth score predicts how likely the response is to initiate a long follow-up discussion thread.

🚀 Quick Start

You can quickly start exploring the capabilities of DialogRPT-depth by clicking on the Colab Notebook Demo.

✨ Features

DialogRPT-depth

Dialog Ranking Pretrained Transformers

How likely a dialog response is upvoted 😊 and/or gets replied 💬?

This is what DialogRPT is designed to predict. It is a set of dialog response ranking models proposed by the Microsoft Research NLP Group. These models are trained on over 100 million human feedback data points. They can be used to enhance existing dialog generation models (e.g., DialoGPT) by re-ranking the generated response candidates.

Quick Links

• EMNLP'20 Paper
• Dataset, training, and evaluation
• Colab Notebook Demo

Model Tasks and Pretrained Models

We have considered the following tasks and provided corresponding pretrained models:

📚 Documentation

Contact

If you have any questions or issues, please create an issue on our repo.

Citation

If you use this work in your research, please cite the following paper:

@inproceedings{gao2020dialogrpt,
    title={Dialogue Response Ranking Training with Large-Scale Human Feedback Data},
    author={Xiang Gao and Yizhe Zhang and Michel Galley and Chris Brockett and Bill Dolan},
    year={2020},
    booktitle={EMNLP}
}