Enhancing Arabic Web Article Classification through Label Noise Reduction and Class-Balanced Training

We present an applied, data-centric study on improving large-scale Arabic web-article classification by detecting and correcting label noise in a 451K-article corpus. Building on a practical AraBERTv02 fine-tuned baseline, we develop an automated noise-detection and relabeling pipeline that combines model confidence, MC-dropout ensemble agreement, and probabilistic flagging inspired by Confident Learning. Using stratified reduced experiments (Train=50k, Val=5k, Test=5k) we automatically relabeled 880 training examples and retrained the classifier. On the held-out 5k test set the cleaned model improved Macro-F1 from 0.9367 (baseline) to 0.9503, a statistically significant gain confirmed by McNemar's test (p ≈ 1.9e-5) and bootstrap 95% CI for Δ Macro-F1 = [0.0082, 0.0190]. We analyze per-class gains, error modes, and cost trade-offs of automated relabeling versus manual review, and release anonymized preprocessing scripts and experiment artifacts to support reproducibility. Our results show that conservative automatic relabelingjudiciously combined with uncertainty estimationcan yield meaningful, reproducible improvements for large-scale Arabic article classification.