Corpus ID: 225040547

ConVEx: Data-Efficient and Few-Shot Slot Labeling

@inproceedings{Henderson2021ConVExDA,
  title={ConVEx: Data-Efficient and Few-Shot Slot Labeling},
  author={Matthew Henderson and Ivan Vuli'c},
  booktitle={NAACL},
  year={2021}
}
We propose ConVEx (Conversational Value Extractor), an efficient pretraining and fine-tuning neural approach for slot-labeling dialog tasks. Instead of relying on more general pretraining objectives from prior work (e.g., language modeling, response selection), ConVEx’s pretraining objective, a novel pairwise cloze task using Reddit data, is well aligned with its intended usage on sequence labeling tasks. This enables learning domain-specific slot labelers by simply fine-tuning decoding layers… Expand

References

SHOWING 1-10 OF 49 REFERENCES
Efficient Intent Detection with Dual Sentence Encoders
ConveRT: Efficient and Accurate Conversational Representations from Transformers
Few-shot Slot Tagging with Collapsed Dependency Transfer and Label-enhanced Task-adaptive Projection Network
Learning to Classify Intents and Slot Labels Given a Handful of Examples
DIET: Lightweight Language Understanding for Dialogue Systems
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Pre-training via Paraphrasing
Few-Shot Question Answering by Pretraining Span Selection
TapNet: Neural Network Augmented with Task-Adaptive Projection for Few-Shot Learning
...
1
2
3
4
5
...