Is BERT Robust to Label Noise? A Study on Learning with Noisy Labels in Text Classification
This work addresses the problem of label noise in NLP for researchers and practitioners, highlighting that current solutions may be inadequate, which is incremental as it builds on prior noise-handling research.
The study investigates whether BERT is robust to label noise in text classification, finding that existing noise-handling methods often fail to improve or even worsen performance across various noise types.
Incorrect labels in training data occur when human annotators make mistakes or when the data is generated via weak or distant supervision. It has been shown that complex noise-handling techniques - by modeling, cleaning or filtering the noisy instances - are required to prevent models from fitting this label noise. However, we show in this work that, for text classification tasks with modern NLP models like BERT, over a variety of noise types, existing noisehandling methods do not always improve its performance, and may even deteriorate it, suggesting the need for further investigation. We also back our observations with a comprehensive analysis.