Dialogue Language Model with Large-Scale Persona Data Engineering
Mengze Hong, Chen Jason Zhang, Chaotao Chen, Rongzhong Lian, Di Jiang
TL;DR
The paper tackles persona inconsistency in open-domain dialogue by leveraging a large-scale persona dialogue dataset built through a summarization-based persona extraction model and a persona augmentation strategy. It introduces PPDS, a Transformer-based model pre-trained on 211M augmented persona-dialogue samples using the UniLM framework, and then fine-tuned on PERSONA-CHAT. Quantitative and human evaluations show that pre-training, augmentation, and fine-tuning collectively yield significant improvements in persona consistency and response quality, outperforming strong baselines like DialoGPT. The work demonstrates a practical pathway to industrially deployable, persona-consistent dialogue systems and outlines future directions including LLM-in-the-loop and multimodal persona data sources.
Abstract
Maintaining persona consistency is paramount in the application of open-domain dialogue systems, as exemplified by models like ChatGPT. Despite significant advancements, the limited scale and diversity of current persona dialogue datasets remain challenges to achieving robust persona-consistent dialogue models. In this study, drawing inspiration from the success of large-scale pre-training, we introduce PPDS, an open-domain persona dialogue system that employs extensive generative pre-training on a persona dialogue dataset to enhance persona consistency. Specifically, we present a persona extraction model designed to autonomously and precisely generate vast persona dialogue datasets. Additionally, we unveil a pioneering persona augmentation technique to address the invalid persona bias inherent in the constructed dataset. Both quantitative and human evaluations consistently highlight the superior response quality and persona consistency of our proposed model, underscoring its effectiveness.