从任务导向对话中无监督发现流程
Unsupervised Flow Discovery from Task-oriented Dialogues
- CISUC, Universidade de Coimbra(科英布拉大学系统与计算机工程中心,科英布拉大学)
- DEI, FCTUC, Universidade de Coimbra(科英布拉大学工程学院,科英布拉大学)
- ISEC, Instituto Politécnico de Coimbra(科英布拉理工学院高等工程学院)
机构由 AI 辅助整理,请以论文原文为准。
中文总结 AI 辅助
本文提出一种从任务导向对话历史中无监督发现对话流程的方法,通过话语向量聚类构建状态转移图,并在MultiWOZ上验证了其提取有意义流程的潜力。
中文摘要 AI 辅助
在开发任务导向对话(TOD)系统时,对话流程的设计是一项关键但耗时的工作。我们提出了一种从对话历史中无监督发现流程的方法,从而使该过程适用于任何具有此类历史的领域。简而言之,将话语表示在向量空间中,并根据其语义相似性进行聚类。聚类可被视为对话状态,然后用作转移图的顶点,以可视化方式表示流程。我们展示了从公共TOD数据集MultiWOZ中发现的流程的具体示例。我们进一步阐述了它们对底层对话的重要性和相关性,并引入了一种自动验证指标用于其评估。实验结果表明,所提出的方法在从任务导向对话中提取有意义的流程方面具有潜力。
英文摘要
The design of dialogue flows is a critical but time-consuming task when developing task-oriented dialogue (TOD) systems. We propose an approach for the unsupervised discovery of flows from dialogue history, thus making the process applicable to any domain for which such an history is available. Briefly, utterances are represented in a vector space and clustered according to their semantic similarity. Clusters, which can be seen as dialogue states, are then used as the vertices of a transition graph for representing the flows visually. We present concrete examples of flows, discovered from MultiWOZ, a public TOD dataset. We further elaborate on their significance and relevance for the underlying conversations and introduce an automatic validation metric for their assessment. Experimental results demonstrate the potential of the proposed approach for extracting meaningful flows from task-oriented conversations.