Metric Learning for Dynamic Text Classification

TL;DR

This paper introduces a metric learning approach for dynamic text classification that adapts to changing label sets without retraining the entire model, improving robustness and performance especially in low-data scenarios.

Contribution

It proposes replacing fixed output layers with a learned metric space for flexible, scalable, and efficient dynamic text classification.

Findings

Robust to label set changes without retraining.

Non-Euclidean metrics improve low-data performance.

Supports incremental addition/removal of labels.

Abstract

Traditional text classifiers are limited to predicting over a fixed set of labels. However, in many real-world applications the label set is frequently changing. For example, in intent classification, new intents may be added over time while others are removed. We propose to address the problem of dynamic text classification by replacing the traditional, fixed-size output layer with a learned, semantically meaningful metric space. Here the distances between textual inputs are optimized to perform nearest-neighbor classification across overlapping label sets. Changing the label set does not involve removing parameters, but rather simply adding or removing support points in the metric space. Then the learned metric can be fine-tuned with only a few additional training examples. We demonstrate that this simple strategy is robust to changes in the label space. Furthermore, our results show that learning a non-Euclidean metric can improve performance in the low data regime, suggesting that further work on metric spaces may benefit low-resource research.