A Deep Reinforcement Learning Chatbot
This is an incremental improvement for chatbot development, aimed at enhancing human-computer interaction in conversational AI.
The authors tackled the problem of building a conversational chatbot by developing MILABOT, a deep reinforcement learning system that uses an ensemble of models to generate responses, and it performed significantly better than many competing 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 template-based models, bag-of-words models, sequence-to-sequence neural network and latent variable neural network 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 many competing systems. Due to its machine learning architecture, the system is likely to improve with additional data.