Pretrained Language Models for Dialogue Generation with Multiple Input Sources

TL;DR

This paper investigates methods to adapt pretrained GPT-2 models for dialogue generation involving multiple input sources, focusing on effective fusion techniques to improve response relevance.

Contribution

It introduces novel fusion methods for multiple input sources in pretrained language models, enhancing dialogue response relevance over simple concatenation or averaging.

Findings

Proper fusion methods outperform simple baselines

Fusion improves relevance with dialogue history

Exploration of multiple attention fusion techniques

Abstract

Large-scale pretrained language models have achieved outstanding performance on natural language understanding tasks. However, it is still under investigating how to apply them to dialogue generation tasks, especially those with responses conditioned on multiple sources. Previous work simply concatenates all input sources or averages information from different input sources. In this work, we study dialogue models with multiple input sources adapted from the pretrained language model GPT2. We explore various methods to fuse multiple separate attention information corresponding to different sources. Our experimental results show that proper fusion methods deliver higher relevance with dialogue history than simple fusion baselines.