Models in the Loop: Aiding Crowdworkers with Generative Annotation Assistants

TL;DR

This paper introduces Generative Annotation Assistants (GAAs), which enhance data collection efficiency and model robustness in question answering tasks by providing real-time suggestions to annotators, achieving significant speed-ups and improved model performance.

Contribution

The work presents GAAs as a novel generator-in-the-loop system that maintains DADC benefits while reducing annotation costs and improving model robustness and accuracy.

Findings

Over 30% increase in annotation speed.

More than 5x improvement in model fooling rates.

Higher downstream QA performance with GAA-assisted data.

Abstract

In Dynamic Adversarial Data Collection (DADC), human annotators are tasked with finding examples that models struggle to predict correctly. Models trained on DADC-collected training data have been shown to be more robust in adversarial and out-of-domain settings, and are considerably harder for humans to fool. However, DADC is more time-consuming than traditional data collection and thus more costly per annotated example. In this work, we examine whether we can maintain the advantages of DADC, without incurring the additional cost. To that end, we introduce Generative Annotation Assistants (GAAs), generator-in-the-loop models that provide real-time suggestions that annotators can either approve, modify, or reject entirely. We collect training datasets in twenty experimental settings and perform a detailed analysis of this approach for the task of extractive question answering (QA) for both standard and adversarial data collection. We demonstrate that GAAs provide significant efficiency benefits with over a 30% annotation speed-up, while leading to over a 5x improvement in model fooling rates. In addition, we find that using GAA-assisted training data leads to higher downstream model performance on a variety of question answering tasks over adversarial data collection.