Learning Visually-Grounded Semantics from Contrastive Adversarial Samples

Haoyue Shi, Jiayuan Mao, Tete Xiao, Yuning Jiang, Jian Sun


Abstract
We study the problem of grounding distributional representations of texts on the visual domain, namely visual-semantic embeddings (VSE for short). Begin with an insightful adversarial attack on VSE embeddings, we show the limitation of current frameworks and image-text datasets (e.g., MS-COCO) both quantitatively and qualitatively. The large gap between the number of possible constitutions of real-world semantics and the size of parallel data, to a large extent, restricts the model to establish a strong link between textual semantics and visual concepts. We alleviate this problem by augmenting the MS-COCO image captioning datasets with textual contrastive adversarial samples. These samples are synthesized using language priors of human and the WordNet knowledge base, and enforce the model to ground learned embeddings to concrete concepts within the image. This simple but powerful technique brings a noticeable improvement over the baselines on a diverse set of downstream tasks, in addition to defending known-type adversarial attacks. Codes are available at https://github.com/ExplorerFreda/VSE-C.
Anthology ID:
C18-1315
Volume:
Proceedings of the 27th International Conference on Computational Linguistics
Month:
August
Year:
2018
Address:
Santa Fe, New Mexico, USA
Venue:
COLING
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
3715–3727
Language:
URL:
https://www.aclweb.org/anthology/C18-1315
DOI:
Bib Export formats:
BibTeX MODS XML EndNote
PDF:
http://aclanthology.lst.uni-saarland.de/C18-1315.pdf