Unsupervised Dialogue Topic Segmentation with Topic-aware Utterance Representation
This work addresses dialogue topic segmentation for natural language processing applications, offering an incremental improvement by better utilizing unlabeled data and combining semantic and coherence aspects.
The paper tackles the problem of unsupervised dialogue topic segmentation by proposing a framework that learns topic-aware utterance representations from unlabeled data, achieving significant performance improvements over strong baselines on benchmark datasets like DialSeg711 and Doc2Dial.
Dialogue Topic Segmentation (DTS) plays an essential role in a variety of dialogue modeling tasks. Previous DTS methods either focus on semantic similarity or dialogue coherence to assess topic similarity for unsupervised dialogue segmentation. However, the topic similarity cannot be fully identified via semantic similarity or dialogue coherence. In addition, the unlabeled dialogue data, which contains useful clues of utterance relationships, remains underexploited. In this paper, we propose a novel unsupervised DTS framework, which learns topic-aware utterance representations from unlabeled dialogue data through neighboring utterance matching and pseudo-segmentation. Extensive experiments on two benchmark datasets (i.e., DialSeg711 and Doc2Dial) demonstrate that our method significantly outperforms the strong baseline methods. For reproducibility, we provide our code and data at:https://github.com/AlibabaResearch/DAMO-ConvAI/tree/main/dial-start.