Correctable-DST: Mitigating Historical Context Mismatch between Training and Inference for Improved Dialogue State Tracking.
Hongyan XieHaoxiang SuShuangyong SongHao HuangBo ZouKun DengJianghua LinZhihui ZhangXiaodong HePublished in: EMNLP (2022)
Keyphrases
- state space
- real time
- context aware
- particle filter
- state vector
- contextual information
- neural network
- state estimation
- context sensitive
- visual tracking
- training set
- supervised learning
- training algorithm
- bayesian networks
- online learning
- semi supervised
- mean shift
- kalman filter
- natural language
- probabilistic inference
- training process
- spoken dialogue systems
- computer vision
- interactive question answering