Label-Agnostic Sequence Labeling by Copying Nearest Neighbors
This work addresses sequence labeling tasks for NLP researchers by offering a label-agnostic approach that improves interpretability and accuracy, though it is incremental as it builds on retrieve-and-edit frameworks.
The paper tackles the problem of structured prediction in sequence labeling by proposing a method that copies labels from retrieved neighbors without explicit editing, achieving accurate performance and enabling transfer to new tasks without retraining.
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.