VD-PCR: Improving visual dialog with pronoun coreference resolution

作者：

Highlights：

• A novel framework VD-PCR to improve visual dialog models with pronoun coreference.

• The joint training helps visual dialog models understand pronouns better.

• The history pruning with pronoun coreference prevents overfitting to dialog history.

• VD-PCR achieves state-of-the-art results on the VisDial dataset.

摘要

•A novel framework VD-PCR to improve visual dialog models with pronoun coreference.•The joint training helps visual dialog models understand pronouns better.•The history pruning with pronoun coreference prevents overfitting to dialog history.•VD-PCR achieves state-of-the-art results on the VisDial dataset.

论文关键词：Vision and language,Visual dialog,Pronoun coreference resolution

论文评审过程：Received 18 April 2021, Revised 4 January 2022, Accepted 14 January 2022, Available online 23 January 2022, Version of Record 23 January 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108540