# A Deep Reinforcement Learning Chatbot

**Authors:** Iulian V. Serban, Chinnadhurai Sankar, Mathieu Germain, Saizheng Zhang, Zhouhan Lin, Sandeep Subramanian, Taesup Kim, Michael Pieper, Sarath Chandar, Nan Rosemary Ke, Sai Mudumba, Alexandre de Brebisson, Jose M. R. Sotelo, Dendi Suhubdy, Vincent Michalski, Alexandre Nguyen, Joelle Pineau, Yoshua Bengio  
**arXiv:** [1709.02349](https://arxiv.org/abs/1709.02349)  
**Date:** 2017-09-07  
**Summary page:** https://www.summarizepaper.com/en/paper/1709.02349v1/

## Key points

- The Montreal Institute for Learning Algorithms (MILA) has developed MILABOT, a deep reinforcement learning chatbot
- MILABOT is designed to engage in small talk conversations with humans using speech and text
- The system utilizes a combination of natural language generation and retrieval models
- Training involved reinforcement learning techniques applied to crowdsourced data and real-world user interactions
- MILABOT outperformed competing systems significantly in evaluations conducted through A/B testing with real-world users
- MILABOT has the potential to improve further as it gains additional data
- The research paper on MILABOT provides detailed information about its development and performance
- Overall, MILABOT represents an advancement in conversational AI technology using deep reinforcement learning.

## Summary

The Montreal Institute for Learning Algorithms (MILA) has developed MILABOT, a deep reinforcement learning chatbot, for the Amazon Alexa Prize competition. MILABOT is designed to engage in small talk conversations with humans using both speech and text. The system utilizes a combination of natural language generation and retrieval models, including template-based models, bag-of-words models, sequence-to-sequence neural network models, and latent variable neural network models. To train MILABOT, reinforcement learning techniques were applied to crowdsourced data and real-world user interactions. This training allowed the system to learn how to select appropriate responses from its ensemble of models. In evaluations conducted through A/B testing with real-world users, MILABOT outperformed competing systems significantly. With its machine learning architecture, MILABOT has the potential to further improve as it gains additional data. The research paper on MILABOT provides detailed information about its development and performance. It is authored by Iulian V. Serban, Chinnadhurai Sankar, Mathieu Germain, Saizheng Zhang, Zhouhan Lin , Sandeep Subramanian , Taesup Kim , Michael Pieper , Sarath Chandar , Nan Rosemary Ke , Sai Mudumba Alexandre de Brebisson Jose M.R Sotelo Dendi Suhubdy Vincent Michalski Alexandre Nguyen Joelle Pineau Yoshua Bengio . Overall ,MILABOT represents an advancement in conversational AI technology and demonstrates the effectiveness of deep reinforcement learning in developing chatbots capable of engaging in natural conversations with humans on various topics.