SALP-CG：面向在线医疗对话数据分类与分级的标准对齐大语言模型流水线
Context-Aware Classification and Grading of Sensitive Information in Online Conversational Health Data
AI总结：
SALP-CG通过统一标准和自动化方法，实现在线医疗对话数据的分类与敏感性分级，提升健康数据治理效率。
摘要：
在线医疗咨询生成大量对话型健康数据，这些数据往往包含受保护的健康信息，需要稳健的方法来分类数据类别并分配风险等级，以符合政策和实践要求。然而，现有方法缺乏统一标准和可靠的自动化方法来满足此类对话健康数据的敏感性分类需求。本研究提出了一种基于大语言模型的提取流水线SALP-CG，用于对在线对话健康数据进行隐私风险的分类和分级。我们根据GB/T 39725-2020制定了健康数据分类和分级规则。结合少样本指导、JSON Schema约束解码和确定性高风险规则，后端无关的提取流水线在不同大语言模型中实现了强大的类别合规性和可靠的敏感性。在MedDialog-CN基准测试中，模型表现出稳健的实体计数、高模式合规性和准确的敏感性分级，而最强的模型在最大级别预测中达到微F1=0.900。按敏感性划分的类别景观显示，等级2-3项占主导地位，结合时可重新识别；等级4-5项较少但危害大。SALP-CG能够可靠地帮助在不同大语言模型中对在线对话健康数据进行分类和敏感性分级，提供了一种实用的健康数据治理方法。代码可在https://github.com/dommii1218/SALP-CG上获得。
英文摘要：
Online medical consultations contain sensitive health information whose privacy implications depend not only on the entities mentioned but also on how those entities are described in context. Existing classification and grading approaches often map health-information entities directly to predefined sensitivity levels, potentially overlooking whether a condition is confirmed, suspected, negated, hypothetical, or merely planned for investigation. In this study, we formulate sensitive-information grading in online medical dialogues as a context-aware evaluation task. We develop a standard-informed operational framework that incorporates assertion status, experiencer, test-result status, and information granularity. We further design a naturalistic evaluation setting together with contrastive cases that minimally alter negation, uncertainty, experiencer, or granularity, and compare large language models under mention-only and full-context conditions. The study aims to quantify the contribution of contextual information to sensitivity grading and to characterize safety-critical over- and under-grading errors. Our framework provides a reproducible basis for evaluating whether LLMs can distinguish sensitive entity mentions from contextually established sensitive disclosures.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
