NoiseBench: Benchmarking the Impact of Real Label Noise on Named Entity Recognition

TL;DR

NoiseBench is a new benchmark for evaluating how real-world label noise affects NER models, revealing that real noise is more challenging than simulated noise and current models struggle with it.

Contribution

This paper introduces NoiseBench, a benchmark with real noise types for NER, and analyzes the impact of real versus simulated noise on model performance.

Findings

Real noise is significantly more challenging than simulated noise.

State-of-the-art noise-robust models underperform on real noise.

Current models fall short of their theoretical upper bounds on noisy data.

Abstract

Available training data for named entity recognition (NER) often contains a significant percentage of incorrect labels for entity types and entity boundaries. Such label noise poses challenges for supervised learning and may significantly deteriorate model quality. To address this, prior work proposed various noise-robust learning approaches capable of learning from data with partially incorrect labels. These approaches are typically evaluated using simulated noise where the labels in a clean dataset are automatically corrupted. However, as we show in this paper, this leads to unrealistic noise that is far easier to handle than real noise caused by human error or semi-automatic annotation. To enable the study of the impact of various types of real noise, we introduce NoiseBench, an NER benchmark consisting of clean training data corrupted with 6 types of real noise, including expert errors, crowdsourcing errors, automatic annotation errors and LLM errors. We present an analysis that shows that real noise is significantly more challenging than simulated noise, and show that current state-of-the-art models for noise-robust learning fall far short of their theoretically achievable upper bound. We release NoiseBench to the research community.