AI summaryⓘ
The authors created a new method called MAAM to better detect harmful language in Chinese, especially when the bias is hidden or depends on context. Their approach focuses on keeping important words related to discrimination and uses extra context such as tone, group identity, and attitude to improve detection. They also built a new dataset, ChLGBT, specifically for Chinese LGBT-related biased language with different levels of bias and emotion. Their method improved prediction accuracy and worked well compared to bigger language models, while being simpler and more stable. This suggests their strategy of focusing on key words and context is a useful and efficient way to understand discrimination in language.
discriminatory language detectionChinese language processingimplicit biascontextual calibrationsemantic anchorsLGBT datasetordinal labelsmodel agnosticBrier scorefew-shot prompting
Abstract
Chinese discriminatory-language detection is challenging because harmful intent is often implicit and context-dependent. We propose MAAM (Myopia--Astigmatism Anchor Mechanism), a lightweight, model-agnostic framework inspired by functional visual blur: rather than preserving every token equally, MAAM retains discrimination-relevant semantic anchors and calibrates them with C--I--S contextual priors (Contextual Tone, Group Identity, and Stance Polarity). We also introduce ChLGBT, to our knowledge the first Chinese LGBT-focused discriminatory-language dataset, with 8,120 manually annotated samples and three ordinal labels: explicit bias, implicit bias, and emotional intensity. Across strong encoder baselines, MAAM improves all three prediction dimensions, with consistent gains in accuracy, F1, Brier score, and expected calibration error. Compared with frontier LLM baselines under zero-shot and few-shot prompting protocols, MAAM remains competitive while offering stronger compactness and stability. These results suggest that interpretable anchor preservation and contextual calibration provide a practical alternative to heavier model scaling for Chinese discriminatory-language assessment.