MuMuQA: Multimedia Multi-Hop News Question Answering via Cross-Media Knowledge Extraction and Grounding

TL;DR

This paper introduces MuMuQA, a new multimedia question answering benchmark involving cross-media reasoning over news articles with images and text, along with a data augmentation framework to improve model training.

Contribution

It presents a novel benchmark for cross-media QA in news, and a data augmentation method for weak supervision, advancing research in multimedia reasoning.

Findings

Models perform well but still lag behind humans.

The benchmark reveals challenges in cross-media grounding.

Data augmentation improves model training effectiveness.

Abstract

Recently, there has been an increasing interest in building question answering (QA) models that reason across multiple modalities, such as text and images. However, QA using images is often limited to just picking the answer from a pre-defined set of options. In addition, images in the real world, especially in news, have objects that are co-referential to the text, with complementary information from both modalities. In this paper, we present a new QA evaluation benchmark with 1,384 questions over news articles that require cross-media grounding of objects in images onto text. Specifically, the task involves multi-hop questions that require reasoning over image-caption pairs to identify the grounded visual object being referred to and then predicting a span from the news body text to answer the question. In addition, we introduce a novel multimedia data augmentation framework, based on cross-media knowledge extraction and synthetic question-answer generation, to automatically augment data that can provide weak supervision for this task. We evaluate both pipeline-based and end-to-end pretraining-based multimedia QA models on our benchmark, and show that they achieve promising performance, while considerably lagging behind human performance hence leaving large room for future work on this challenging new task.