BROWSE
Detail
Conference Paper
Dialogue State Tracking with Zero-Shot and Few-Shot Learning for Generalization: A Review
Cited 0 time in 
Share 
Share
- Authors
- Seungyeon Kim, Yejin Park, Junseong Bang
- Issue Date
- 2022-08
- Citation
- International Conference on Platform Technology and Service (PlatCon) 2022, pp.75-79
- Language
- English
- Type
- Conference Paper
- Abstract
- Research on Dialogue State Tracking (DST) has achieved meaningful advancements in benchmark datasets. However, the generalization ability of DST models to handle unseen data robustly remains an issue. Hence, recent studies on DST with zero-shot and few-shot learning are reviewed in this paper. For a task-oriented dialogue system, DST is explained by introducing datasets and evaluation metrics. DST models could be categorized into four groups: DST based on a pre-trained model, DST using a description, DST using a prompt, and DST with cross-task. Characteristics of each model are described and the performance of the model experimented under the same conditions is summarized.
- KSP Keywords
- Benchmark datasets, Generalization ability, Pre-trained model, State tracking, Task-oriented, Zero-shot, dialogue system, evaluation metrics