# The RLLChatbot: a solution to the ConvAI Challenge

**Authors:** Nicolas A. Gontier, Koustuv Sinha, Peter Henderson, Iulian Serban, Michael Noseworthy, Prasanna Parthasarathi, Joelle Pineau  
**arXiv:** [1811.02714](https://arxiv.org/abs/1811.02714)  
**Date:** 2018-11-07  
**Summary page:** https://www.summarizepaper.com/en/paper/1811.02714v1/

## Key points

- RLLChatbot participated in the 2017 ConvAI challenge
- Goal of the challenge is to develop conversational systems for coherent and open-ended conversations
- RLLChatbot uses deep learning and reinforcement learning tools
- Thorough description of how the dialog system was built and trained using public-domain datasets
- Introduction of ensemble model, message ranking, and selection methods
- New open-source conversational dataset improves ranking and selection mechanisms
- Challenges faced by current conversational systems in maintaining coherence highlighted
- Importance of competitions like ConvAI in driving research development emphasized
- RLLChatbot showcases application of deep learning and reinforcement learning techniques
- Valuable insights into building effective dialog systems for open domain conversations provided

## Summary

The article discusses the RLLChatbot, which participated in the 2017 Conversational Intelligence (ConvAI) challenge. The goal of the challenge is to develop conversational systems that can engage in coherent and open-ended conversations about specific topics. The RLLChatbot aims to understand how deep learning and reinforcement learning tools can be used to build a robust and flexible open domain conversational agent. The authors provide a thorough description of how the dialog system was built and trained using mostly public-domain datasets. They also introduce an ensemble model that incorporates different text generation models, as well as novel message ranking and selection methods. Additionally, they present a new open-source conversational dataset that significantly improves the ranking and selection mechanisms compared to their baseline model. The paper highlights the challenges faced by current conversational systems in maintaining coherent conversations and emphasizes the importance of competitions like ConvAI in driving research development in this area. The RLLChatbot serves as an example of how deep learning and reinforcement learning techniques can be applied to enhance conversational agents' capabilities. Overall, this article provides valuable insights into building effective dialog systems for open domain conversations, offering potential advancements for future research in conversational intelligence.