DialoGPT: Large-Scale Generative Pre-training for Conversational Response Generation

원문 초록

We present a large, tunable neural conversational response generation model, DialoGPT (dialogue generative pre-trained transformer). Trained on 147M conversation-like exchanges extracted from Reddit comment chains over a period spanning from 2005 through 2017, DialoGPT extends the Hugging Face PyTorch transformer to attain a performance close to human both in terms of automatic and human evaluation in single-turn dialogue settings. We show that conversational systems that leverage DialoGPT generate more relevant, contentful and context-consistent responses than strong baseline systems. The pre-trained model and training pipeline are publicly released to facilitate research into neural response generation and the development of more intelligent open-domain dialogue systems.