Cascading versus Joint Modeling for Hierarchical Offensive Language Detection

2026-07-18

AI Summary

This study investigates the trade-offs between cascaded modeling and joint multi-task modeling for fine-grained offensive language detection, focusing on accuracy, parameter count, and inference latency, while optimizing class-imbalance handling strategies for each subtask. We construct a three-stage cascade system with tailored training protocols per stage and employ ablation studies to identify the optimal imbalance mitigation approach. A shared-encoder joint model serves as a baseline, enabling the first controlled quantitative comparison between the two paradigms. Results show that the cascade achieves macro F1 scores of 0.795, 0.716, and 0.557 across the three subtasks, outperforming the joint model by 7.1 points on the most imbalanced task, albeit with three times more parameters and 1.67× higher latency. Ablation analysis further reveals that approximately 20% of cascade errors originate in the first stage and are irrecoverable, underscoring the critical impact of pipeline design.

Abstract

Fine-grained offensive language detection organizes labels into a hierarchical structure, for which two modeling paradigms exist: cascaded decomposition and joint multi-task modeling. Prior work rarely provides a direct, controlled comparison of the two paradigms in terms of accuracy, parameter count, and inference latency, and rarely verifies whether a chosen class-imbalance handling strategy is actually optimal. This paper proposes a three-level cascaded detection system whose training strategy is customized per subtask, together with two verification mechanisms. First, a controlled ablation study determines the best class-imbalance handling strategy for each subtask. Second, a joint multi-task model with a shared encoder is trained as an architectural control, yielding real measurements along the dimensions of accuracy, parameter count, and inference latency. Experiments show that the cascaded system attains macro-F1 scores of 0.795, 0.716, and 0.557 on the three subtasks of the official test set. The ablation study reveals that configuring the loss function purely by imbalance-severity intuition is suboptimal; reconfiguring based on the ablation results improves both performance and stability. End-to-end cascade evaluation shows that roughly one-fifth of the errors in the cascade pipeline originate from the first-stage filter and cannot be corrected by subsequent stages. Relative to the joint multi-task model, the cascaded architecture achieves higher accuracy on all three subtasks, with a 7.1-point macro-F1 gain on the most severely imbalanced subtask, at the cost of three times the parameters and 1.67 times the inference latency. Together, these results establish an explicit, quantifiable trade-off between the accuracy advantage of cascaded architectures and their deployment cost.

Problem

• hierarchical offensive language detection
• cascaded modeling
• joint multi-task modeling
• class imbalance
• model comparison

Innovation

• cascaded modeling
• joint multi-task learning
• class imbalance
• hierarchical offensive language detection
• controlled ablation study