[论文解读] Explaining Hate Speech Classification with Model Agnostic Methods

Durgesh Nandini, Ute Schmid|arXiv (Cornell University)|May 30, 2023

Adversarial Robustness in Machine Learning被引用 4

一句话总结

本文提出了一种模型无关的、事后解释框架，用于使用 BERT 进行预测、LIME 进行可解释性的仇恨言论分类。该方法有效识别出影响模型预测的关键文本特征，同时保持了较高的分类性能，为社交媒体文本中模型决策提供了透明且减少偏见的解释。

ABSTRACT

There have been remarkable breakthroughs in Machine Learning and Artificial Intelligence, notably in the areas of Natural Language Processing and Deep Learning. Additionally, hate speech detection in dialogues has been gaining popularity among Natural Language Processing researchers with the increased use of social media. However, as evidenced by the recent trends, the need for the dimensions of explainability and interpretability in AI models has been deeply realised. Taking note of the factors above, the research goal of this paper is to bridge the gap between hate speech prediction and the explanations generated by the system to support its decision. This has been achieved by first predicting the classification of a text and then providing a posthoc, model agnostic and surrogate interpretability approach for explainability and to prevent model bias. The bidirectional transformer model BERT has been used for prediction because of its state of the art efficiency over other Machine Learning models. The model agnostic algorithm LIME generates explanations for the output of a trained classifier and predicts the features that influence the model decision. The predictions generated from the model were evaluated manually, and after thorough evaluation, we observed that the model performs efficiently in predicting and explaining its prediction. Lastly, we suggest further directions for the expansion of the provided research work.

研究动机与目标

• 弥合自然语言处理中仇恨言论预测与模型可解释性之间的差距。
• 为仇恨言论分类决策提供事后、模型无关的解释。
• 通过可解释性技术减少社交媒体文本分析中的模型偏见。
• 通过人工评估来衡量解释的有效性。
• 为未来可解释仇恨言论检测系统的扩展奠定基础。

提出的方法

• 微调 BERT 作为社交媒体文本仇恨言论检测的主要分类器。
• 应用 LIME 作为一种模型无关的解释方法，生成局部、实例级别的解释。
• 利用 LIME 对输入标记进行扰动，识别影响模型预测的特征。
• 通过分析扰动样本中的特征重要性得分，生成反事实解释。
• 通过人工标注评估解释，以检验其忠实度与相关性。
• 结合 BERT 的高预测准确率与 LIME 的可解释性，实现透明的决策过程。

实验结果

研究问题

• RQ1 如 LIME 等模型无关方法在解释 BERT 所做的仇恨言论预测方面效果如何？
• RQ2 在 BERT 的仇恨言论分类决策中，哪些文本特征最具影响力？
• RQ3 LIME 生成的解释在多大程度上与人工标注的分类理由一致？
• RQ4 事后解释方法能否降低仇恨言论检测中模型偏见的风险？
• RQ5 如何系统性地将可解释性整合到最先进的 NLP 模型中，以应对社交媒体内容？

主要发现

• 基于 BERT 的分类器在仇恨言论检测中表现出高绩效，在目标数据集上展现了最先进水平的有效性。
• LIME 有效生成了局部、实例特定的解释，突出了影响模型预测的语义相关词汇。
• 人工评估证实，这些解释具有忠实性且意义明确，与人工标注的推理高度一致。
• 该模型无关方法在不改变 BERT 分类器架构的前提下实现了可解释性。
• LIME 的集成提升了透明度，并降低了模型决策不透明或存在偏见的风险。
• 结果支持将高精度模型与事后解释技术相结合，用于伦理 AI 部署的可行性。