Towards Neural Speaker Modeling in Multi-Party Conversation: The Task, Dataset, and Models
This work addresses a gap in neural dialog systems for researchers, but it is incremental as it builds on existing speaker modeling concepts with new tasks and data.
The paper tackles the lack of established tasks and datasets for neural speaker modeling in multi-party conversation by proposing speaker classification as a surrogate task and collecting a large dataset, with experiments showing that hybrid models outperform single-component ones in this classification.
Neural network-based dialog systems are attracting increasing attention in both academia and industry. Recently, researchers have begun to realize the importance of speaker modeling in neural dialog systems, but there lacks established tasks and datasets. In this paper, we propose speaker classification as a surrogate task for general speaker modeling, and collect massive data to facilitate research in this direction. We further investigate temporal-based and content-based models of speakers, and propose several hybrids of them. Experiments show that speaker classification is feasible, and that hybrid models outperform each single component.