Cross-modal Retrieval for Knowledge-based Visual Question Answering

Paul Lerner (1) , Olivier Ferret (2) , Camille Guinaudeau (3)

1 MLIA - Machine Learning and Information Access
2 DIASI (CEA, LIST) - Département Intelligence Ambiante et Systèmes Interactifs
3 STL - Sciences et Technologies des Langues - LISN

Knowledge-based Visual Question Answering about Named Entities is a challenging task that requires retrieving information from a multimodal Knowledge Base. Named entities have diverse visual representations and are therefore difficult to recognize. We argue that cross-modal retrieval may help bridge the semantic gap between an entity and its depictions, and is foremost complementary with mono-modal retrieval. We provide empirical evidence through experiments with a multimodal dual encoder, namely CLIP, on the recent ViQuAE, InfoSeek, and Encyclopedic-VQA datasets. Additionally, we study three different strategies to fine-tune such a model: mono-modal, cross-modal, or joint training. Our method, which combines mono-and cross-modal retrieval, is competitive with billion-parameter models on the three datasets, while being conceptually simpler and computationally cheaper.

Mots clés: Visual Question Answering; Multimodal; Cross-modal Retrieval; Named Entities

Domaines: Recherche d'information [cs.IR]; Informatique et langage [cs.CL]

Paul Lerner, Olivier Ferret, Camille Guinaudeau. Cross-modal Retrieval for Knowledge-based Visual Question Answering. 46th European Conference on Information Retrieval (ECIR 2024), 2024, Glasgow, United Kingdom. ⟨hal-04384431⟩