Spoken ObjectNet: A Bias-Controlled Spoken Caption Dataset

TL;DR

Spoken ObjectNet is a bias-controlled spoken caption dataset designed to evaluate and improve the real-world performance of cross-modal models by removing biases present in existing datasets.

Contribution

The paper introduces Spoken ObjectNet, a new bias-controlled spoken caption dataset with improved data collection methods and baseline evaluations for cross-modal tasks.

Findings

Models trained on other datasets perform poorly on Spoken ObjectNet due to biases.

Bias controls in the dataset effectively reveal limitations of existing models.

Baseline results demonstrate the dataset's utility for evaluating spoken language models.

Abstract

Visually-grounded spoken language datasets can enable models to learn cross-modal correspondences with very weak supervision. However, modern audio-visual datasets contain biases that undermine the real-world performance of models trained on that data. We introduce Spoken ObjectNet, which is designed to remove some of these biases and provide a way to better evaluate how effectively models will perform in real-world scenarios. This dataset expands upon ObjectNet, which is a bias-controlled image dataset that features similar image classes to those present in ImageNet. We detail our data collection pipeline, which features several methods to improve caption quality, including automated language model checks. Lastly, we show baseline results on image retrieval and audio retrieval tasks. These results show that models trained on other datasets and then evaluated on Spoken ObjectNet tend to perform poorly due to biases in other datasets that the models have learned. We also show evidence that the performance decrease is due to the dataset controls, and not the transfer setting.