Pretrained Language Models for Dialogue Generation with Multiple Input Sources
This work addresses a specific challenge in dialogue systems for AI applications, but it is incremental as it builds on existing models like GPT2.
The paper tackled the problem of applying pretrained language models to dialogue generation with multiple input sources, finding that proper fusion methods improved relevance to dialogue history compared to simple baselines.
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.