A Deep Reinforcement Learning Chatbot (Short Version)
This work addresses the challenge of building effective chatbots for real-world applications like the Amazon Alexa Prize, though it is incremental in combining existing techniques.
The researchers tackled the problem of developing an open-domain conversational agent by creating MILABOT, a deep reinforcement learning chatbot that uses an ensemble of models and was trained on crowdsourced and real-world data. It performed significantly better than other systems in A/B testing with real-world users.
We present MILABOT: a deep reinforcement learning chatbot developed by the Montreal Institute for Learning Algorithms (MILA) for the Amazon Alexa Prize competition. MILABOT is capable of conversing with humans on popular small talk topics through both speech and text. The system consists of an ensemble of natural language generation and retrieval models, including neural network and template-based models. By applying reinforcement learning to crowdsourced data and real-world user interactions, the system has been trained to select an appropriate response from the models in its ensemble. The system has been evaluated through A/B testing with real-world users, where it performed significantly better than other systems. The results highlight the potential of coupling ensemble systems with deep reinforcement learning as a fruitful path for developing real-world, open-domain conversational agents.