Label-Agnostic Sequence Labeling by Copying Nearest Neighbors

TL;DR

This paper introduces a label-agnostic copying method for sequence labeling that leverages retrieved neighbors, enabling accurate, interpretable predictions across tasks without retraining.

Contribution

It presents a novel explicit copying approach for sequence labeling that is label-agnostic and adaptable to new tasks without retraining.

Findings

High accuracy in sequence labeling through copying from neighbors

Effective transfer to new tasks without retraining

Improved interpretability with dynamic programming control

Abstract

Retrieve-and-edit based approaches to structured prediction, where structures associated with retrieved neighbors are edited to form new structures, have recently attracted increased interest. However, much recent work merely conditions on retrieved structures (e.g., in a sequence-to-sequence framework), rather than explicitly manipulating them. We show we can perform accurate sequence labeling by explicitly (and only) copying labels from retrieved neighbors. Moreover, because this copying is label-agnostic, we can achieve impressive performance when transferring to new sequence-labeling tasks without retraining. We additionally consider a dynamic programming approach to sequence labeling in the presence of retrieved neighbors, which allows for controlling the number of distinct (copied) segments used to form a prediction, and leads to both more interpretable and accurate predictions.