HNC: Leveraging Hard Negative Captions towards Models with Fine-Grained Visual-Linguistic Comprehension Capabilities

Esra Dönmez, Pascal Tilli, Hsiu-Yu Yang, Ngoc Thang Vu, Carina Silberer


Abstract
Image-Text-Matching (ITM) is one of the defacto methods of learning generalized representations from a large corpus in Vision and Language (VL). However, due to the weak association between the web-collected image–text pairs, models fail to show fine-grained understanding of the combined semantics of these modalities. To this end, we propose Hard Negative Captions (HNC): an automatically created dataset containing foiled hard negative captions for ITM training towards achieving fine-grained cross-modal comprehension in VL. Additionally, we provide a challenging manually-created test set for benchmarking models on a fine-grained cross-modal mismatch with varying levels of compositional complexity. Our results show the effectiveness of training on HNC by improving the models’ zero-shot capabilities in detecting mismatches on diagnostic tasks and performing robustly under noisy visual input scenarios. Also, we demonstrate that HNC models yield a comparable or better initialization for fine-tuning. Our code and data are publicly available.
Anthology ID:
2023.conll-1.24
Volume:
Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL)
Month:
December
Year:
2023
Address:
Singapore
Editors:
Jing Jiang, David Reitter, Shumin Deng
Venue:
CoNLL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
364–388
Language:
URL:
https://aclanthology.org/2023.conll-1.24
DOI:
10.18653/v1/2023.conll-1.24
Bibkey:
Cite (ACL):
Esra Dönmez, Pascal Tilli, Hsiu-Yu Yang, Ngoc Thang Vu, and Carina Silberer. 2023. HNC: Leveraging Hard Negative Captions towards Models with Fine-Grained Visual-Linguistic Comprehension Capabilities. In Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL), pages 364–388, Singapore. Association for Computational Linguistics.
Cite (Informal):
HNC: Leveraging Hard Negative Captions towards Models with Fine-Grained Visual-Linguistic Comprehension Capabilities (Dönmez et al., CoNLL 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.conll-1.24.pdf
Video:
 https://aclanthology.org/2023.conll-1.24.mp4