Automated Speaking Assessment (ASA) plays a crucial role in evaluating second-language (L2) learners proficiency. However, ASA models often suffer from class imbalance, leading to biased predictions. To address this, we introduce a novel objective for training ASA models, dubbed the Balancing Logit Variation (BLV) loss, which perturbs model predictions to improve feature representation for minority classes without modifying the dataset. Evaluations on the ICNALE benchmark dataset show that integrating the BLV loss into a celebrated text-based (BERT) model significantly enhances classification accuracy and fairness, making automated speech evaluation more robust for diverse learners.
@article{arxiv.2509.03010,
title = {Mitigating Data Imbalance in Automated Speaking Assessment},
author = {Fong-Chun Tsai and Kuan-Tang Huang and Bi-Cheng Yan and Tien-Hong Lo and Berlin Chen},
journal= {arXiv preprint arXiv:2509.03010},
year = {2026}
}
Comments
Accepted by APSIPA 2025; revised figure, references added