A Thousand Words Are Worth More Than a Picture: Natural Language-Centric Outside-Knowledge Visual Question Answering

TL;DR

This paper introduces TRiG, a novel framework for outside-knowledge visual question answering that converts images into text to leverage large-scale knowledge bases and language models, significantly improving performance.

Contribution

The paper proposes a paradigm shift transforming images into text for better external knowledge integration in VQA, and introduces the TRiG framework that outperforms existing methods.

Findings

TRiG outperforms state-of-the-art methods by at least 11.1%

Transforming images into text enables better external knowledge utilization

The framework is flexible and compatible with various models and knowledge bases

Abstract

Outside-knowledge visual question answering (OK-VQA) requires the agent to comprehend the image, make use of relevant knowledge from the entire web, and digest all the information to answer the question. Most previous works address the problem by first fusing the image and question in the multi-modal space, which is inflexible for further fusion with a vast amount of external knowledge. In this paper, we call for a paradigm shift for the OK-VQA task, which transforms the image into plain text, so that we can enable knowledge passage retrieval, and generative question-answering in the natural language space. This paradigm takes advantage of the sheer volume of gigantic knowledge bases and the richness of pre-trained language models. A Transform-Retrieve-Generate framework (TRiG) framework is proposed, which can be plug-and-played with alternative image-to-text models and textual knowledge bases. Experimental results show that our TRiG framework outperforms all state-of-the-art supervised methods by at least 11.1% absolute margin.