Improving Conditioning in Context-Aware Sequence to Sequence Models
This work addresses a specific bottleneck in context-aware generation for applications like document translation, but it is incremental as it builds on standard sequence-to-sequence models.
The paper tackled the problem of generating sequences conditioned on both a short query and a long context, such as in abstractive question answering, by modifying sequence-to-sequence models to intertwine query and context attention and adding data augmentation. Experiments on three tasks showed consistent improvements.
Neural sequence to sequence models are well established for applications which can be cast as mapping a single input sequence into a single output sequence. In this work, we focus on cases where generation is conditioned on both a short query and a long context, such as abstractive question answering or document-level translation. We modify the standard sequence-to-sequence approach to make better use of both the query and the context by expanding the conditioning mechanism to intertwine query and context attention. We also introduce a simple and efficient data augmentation method for the proposed model. Experiments on three different tasks show that both changes lead to consistent improvements.