Deep Reinforcement Learning for Dialogue Generation

TLDR

Recent neural dialogue models promise high‑quality responses but are shortsighted, ignoring future impact, so modeling dialogue trajectories with reinforcement learning is essential for coherence and engagement. This study demonstrates how to integrate deep reinforcement learning to model future reward in chatbot dialogue. The authors simulate conversations between two virtual agents and use policy‑gradient methods to reward sequences that are informative, coherent, and easy to answer. Evaluation on diversity, length, and human judgments shows the algorithm generates more interactive responses and sustains longer conversations, marking a first step toward long‑term success in neural dialogue.

Abstract

Recent neural models of dialogue generation offer great promise for generating responses for conversational agents, but tend to be shortsighted, predicting utterances one at a time while ignoring their influence on future outcomes. Modeling the future direction of a dialogue is crucial to generating coherent, interesting dialogues, a need which led traditional NLP models of dialogue to draw on reinforcement learning. In this paper, we show how to integrate these goals, applying deep reinforcement learning to model future reward in chatbot dialogue. The model simulates dialogues between two virtual agents, using policy gradient methods to reward sequences that display three useful conversational properties: informativity (non-repetitive turns), coherence, and ease of answering (related to forward-looking function). We evaluate our model on diversity, length as well as with human judges, showing that the proposed algorithm generates more interactive responses and manages to foster a more sustained conversation in dialogue simulation. This work marks a first step towards learning a neural conversational model based on the long-term success of dialogues.