2025 Volume 32 Issue 3 Pages 829-858
This work investigates the impact of data augmentation on uncertainty estimation in Named Entity Recognition (NER) tasks. For the future advance of NER in safety-critical fields like healthcare and finance, it is essential to achieve accurate predictions with calibrated confidence when applying Deep Neural Networks (DNNs), including Pre-trained Language Models, as a real-world application. However, DNNs are prone to miscalibration, which limits their applicability. Moreover, existing methods for calibration and uncertainty estimation are computational expensive. Our investigation in NER found that data augmentation improves calibration and uncertainty in cross-genre and cross-lingual setting, especially in-domain setting. We also observed that increasing the data augmentation size further improved the uncertainty estimation performance in NER. In addition, its tendency suggests that the improvement correlated with the perplexity of sentences generated through data augmentation methods.
