ECOL-R: Encouraging Copying in Novel Object Captioning with Reinforcement Learning

Yufei Wang, Ian Wood, Stephen Wan, Mark Johnson


Abstract
Novel Object Captioning is a zero-shot Image Captioning task requiring describing objects not seen in the training captions, but for which information is available from external object detectors. The key challenge is to select and describe all salient detected novel objects in the input images. In this paper, we focus on this challenge and propose the ECOL-R model (Encouraging Copying of Object Labels with Reinforced Learning), a copy-augmented transformer model that is encouraged to accurately describe the novel object labels. This is achieved via a specialised reward function in the SCST reinforcement learning framework (Rennie et al., 2017) that encourages novel object mentions while maintaining the caption quality. We further restrict the SCST training to the images where detected objects are mentioned in reference captions to train the ECOL-R model. We additionally improve our copy mechanism via Abstract Labels, which transfer knowledge from known to novel object types, and a Morphological Selector, which determines the appropriate inflected forms of novel object labels. The resulting model sets new state-of-the-art on the nocaps (Agrawal et al., 2019) and held-out COCO (Hendricks et al., 2016) benchmarks.
Anthology ID:
2021.eacl-main.104
Volume:
Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume
Month:
April
Year:
2021
Address:
Online
Editors:
Paola Merlo, Jorg Tiedemann, Reut Tsarfaty
Venue:
EACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
1222–1234
Language:
URL:
https://aclanthology.org/2021.eacl-main.104
DOI:
10.18653/v1/2021.eacl-main.104
Bibkey:
Cite (ACL):
Yufei Wang, Ian Wood, Stephen Wan, and Mark Johnson. 2021. ECOL-R: Encouraging Copying in Novel Object Captioning with Reinforcement Learning. In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume, pages 1222–1234, Online. Association for Computational Linguistics.
Cite (Informal):
ECOL-R: Encouraging Copying in Novel Object Captioning with Reinforcement Learning (Wang et al., EACL 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.eacl-main.104.pdf
Data
MS COCONoCapsOpen Images V4