Skip to content

malihealikhani/Cross-modal_Coherence_Modeling

Repository files navigation

Cross-modal Coherence Modeling for Caption Generation

Malihe Alikhani, Piyush Sharma, Shengjie Li, Radu Soricut, Matthew Stone

We use coherence relations inspired by computational models of discourse to study the information needs and goals of image captioning. Using an annotation protocol specifically devised for capturing image--caption coherence relations, we annotate 10,000 instances from publicly-available image--caption pairs. We show that these coherence annotations can be exploited to learn relation classifiers as an intermediary step, and also train coherence-aware, controllable image captioning models. The results show a dramatic improvement in the consistency and quality of the generated captions with respect to information needs specified via coherence relations.

About

Cross-modal Coherence Modeling for Caption Generation

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published