Visual dialogue without vision or dialogue

Conference item

We characterise some of the quirks and shortcomings in the exploration of visual dialogue (VD)—a sequential question-answering task where the questions and corresponding answers are related through given visual stimuli. To do so, we develop an embarrassingly simple method based on canonical correlation analysis (CCA) that, on the standard dataset, achieves near state-of-the-art performance on mean rank (MR). In direct contrast to current complex and over-parametrised architectures that are both compute and time intensive, our method ignores the visual stimuli, ignores the sequencing of dialogue, does not need gradients, uses off-the-shelf feature extractors, has at least an order of magnitude fewer parameters, and learns in practically no time. We argue that these results are indicative of issues in current approaches to visual dialogue and conduct analyses to highlight implicit dataset biases and effects of over-constrained evaluation metrics. Our code is publicly available.

Authors: Massiceti, D; Dokania, P; Siddharth, N; Torr, P

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Neural Information Processing Systems Foundation
• Publication date: 2018-12-07
• Acceptance date: 2018-11-10
• Event title: NeurIPS 2018 Workshop on Critiquing and Correcting Trends in Machine Learning
• Event location: Montreal, Canada
• Event website: https://nips.cc/Conferences/2018/Schedule?showEvent=10911
• Event start date: 2018-12-07
• Event end date: 2018-12-07
• Language: English