Changing the World by Changing the Data

TL;DR

This paper discusses the importance and impact of dataset curation in NLP, arguing that data design influences model behavior and societal biases, and that careful curation is inevitable and impactful.

Contribution

It provides a comprehensive analysis of data curation's role in NLP, emphasizing its significance over model development and highlighting the need for thoughtful dataset design.

Findings

Data curation is already prevalent and influential.

Dataset design affects model biases and societal impacts.

Thoughtful curation can shape the future of NLP models.

Abstract

NLP community is currently investing a lot more research and resources into development of deep learning models than training data. While we have made a lot of progress, it is now clear that our models learn all kinds of spurious patterns, social biases, and annotation artifacts. Algorithmic solutions have so far had limited success. An alternative that is being actively discussed is more careful design of datasets so as to deliver specific signals. This position paper maps out the arguments for and against data curation, and argues that fundamentally the point is moot: curation already is and will be happening, and it is changing the world. The question is only how much thought we want to invest into that process.