多视角融合的无监督对话主题分割模型

打开文本图片集

中图分类号：TP391.1 文献标志码：A 文章编号： 1000-5013（2026）02-0183-10

Abstract：An unsupervised dialogue topic segmentation model based on multi-view fusion is proposed in this paper，which jointly models semantic similarity，logical coherence and summary consistency. The proposed framework adaptively integrates information from multiple perspectives through a hybrid mechanism combining static weighting with dynamic gating. Furthermore，a unified optimization objective is established by combining neighboring uterance matching loss，summary consistency loss，and semantic-correlation modeling loss. Experimental results on three representative datasets show that the proposed model consistently achieves supe rior performance， effectively improving robustness and global semantic coherence.

Keywords：dialogue topic segmentation； dialogue summarization； topic modeling；unsupervised； neighboringutterance matching

对话主题分割（DTS）通过将对话分割成主题连贯的部分，以揭示对话的主题结构[1]，对话主题分割对各种下游对话相关的自然语言处理（NLP）任务起着至关重要的作用。（剩余16897字）