Mitigating Data Imbalance in Automated Speaking Assessment

2025/09/03 by Tsai, Fong-Chun, Huang, Kuan-Tang, Yan, Bi-Cheng +2
#Audio and Speech Processing (eess.AS) #Computation and Language (cs.CL) #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #electronic engineering #information engineering

paper · doi:10.48550/arxiv.2509.03010

Abstract

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.

Citations

• An Effective Automated Speaking Assessment Approach to Mitigating Data Scarcity and Imbalanced Distribution
• Balancing Logit Variation for Long-tailed Semantic Segmentation
• Robust Speech Recognition via Large-Scale Weak Supervision
• L2 proficiency assessment using self-supervised speech representations
• Proficiency assessment of L2 spoken English using wav2vec 2.0
• data2vec: A General Framework for Self-supervised Learning in Speech, Vision and Language
• HuBERT: Self-Supervised Speech Representation Learning by Masked\n Prediction of Hidden Units
• wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations
• A systematic study of the class imbalance problem in convolutional neural networks
• Attention-Based Models for Speech Recognition

Related

• An Effective Automated Speaking Assessment Approach to Mitigating Data Scarcity and Imbalanced Distribution
• A Novel Data Augmentation Approach for Automatic Speaking Assessment on Opinion Expressions
• One Whisper to Grade Them All
• An Effective Strategy for Modeling Score Ordinality and Non-uniform Intervals in Automated Speaking Assessment
• The NTNU System at the S&I Challenge 2025 SLA Open Track