C2C-GenDA: Cluster-to-Cluster Generation For Data Augmentation Of Slot Filling

First, as shown in Figure 6, every slot tag was reformulated as a pure language Question. Goel et al. (2018) create two sorts of candidate sets: N-gram candidate set and spoken language understanding (SLU) candidate set. 2018) by exploiting the language illustration energy of BERT to enhance the generalization capability. At the middle of the…