Uncertainty-Aware Label Refinement for Sequence Labeling

TL;DR

This paper presents a novel two-stage, uncertainty-aware label refinement framework for sequence labeling that models long-term dependencies more efficiently and improves accuracy over traditional CRF methods.

Contribution

Introduces a two-stage decoding framework with Bayesian neural networks for uncertainty estimation, enabling faster and more accurate sequence labeling.

Findings

Outperforms CRF-based methods on benchmark datasets

Achieves faster inference through parallel decoding

Effectively models long-term label dependencies

Abstract

Conditional random fields (CRF) for label decoding has become ubiquitous in sequence labeling tasks. However, the local label dependencies and inefficient Viterbi decoding have always been a problem to be solved. In this work, we introduce a novel two-stage label decoding framework to model long-term label dependencies, while being much more computationally efficient. A base model first predicts draft labels, and then a novel two-stream self-attention model makes refinements on these draft predictions based on long-range label dependencies, which can achieve parallel decoding for a faster prediction. In addition, in order to mitigate the side effects of incorrect draft labels, Bayesian neural networks are used to indicate the labels with a high probability of being wrong, which can greatly assist in preventing error propagation. The experimental results on three sequence labeling benchmarks demonstrated that the proposed method not only outperformed the CRF-based methods but also greatly accelerated the inference process.