利用弱监督信号修复不安全的对话回复
Healing Unsafe Dialogue Responses with Weak Supervision Signals
AI总结：
本文提出无监督伪标签采样方法 TEMP，通过聚类与自适应锐化采样自动生成安全回复，在弱监督和无监督对话场景下均取得领先效果。
摘要：
近年来，大规模对话系统生成不安全回复的问题日益受到关注，智能体会从真实世界语料中学习到冒犯性或有偏见的言行。已有一些方法通过流水线方式检测并替换不安全的训练样本来解决上述问题。这些方法虽然有效，但标注成本高昂，且难以适应未知场景和对抗攻击。此外，忽略提供安全回复（例如简单地用模板替换）会导致对话信息缺失的问题。为解决这些问题，我们提出了一种无监督伪标签采样方法 TEMP，能够自动分配潜在的安全回复。具体而言，我们的 TEMP 方法将回复聚类为若干簇，并采用自适应锐化的采样策略为每个样本采样多个标签，这一设计受到如下观察的启发：簇中的不安全样本通常数量很少，且分布在尾部。在闲聊和任务导向对话上的大量实验表明，我们的 TEMP 方法在弱监督信号下优于最先进模型，并在无监督学习设置下取得具有可比性的结果。
英文摘要：
Recent years have seen increasing concerns about the unsafe response generation of large-scale dialogue systems, where agents will learn offensive or biased behaviors from the real-world corpus. Some methods are proposed to address the above issue by detecting and replacing unsafe training examples in a pipeline style. Though effective, they suffer from a high annotation cost and adapt poorly to unseen scenarios as well as adversarial attacks. Besides, the neglect of providing safe responses (e.g. simply replacing with templates) will cause the information-missing problem of dialogues. To address these issues, we propose an unsupervised pseudo-label sampling method, TEMP, that can automatically assign potential safe responses. Specifically, our TEMP method groups responses into several clusters and samples multiple labels with an adaptively sharpened sampling strategy, inspired by the observation that unsafe samples in the clusters are usually few and distribute in the tail. Extensive experiments in chitchat and task-oriented dialogues show that our TEMP outperforms state-of-the-art models with weak supervision signals and obtains comparable results under unsupervised learning settings.
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。
