• Publications
  • Influence
Few-shot Pseudo-Labeling for Intent Detection
A folding/unfolding hierarchical clustering algorithm which assigns weighted pseudo-labels to unlabeled user utterances for few-shot intent detection yields significant improvement over existing solutions, thereby providing an even stronger state-of-the-art ensemble method. Expand
A Neural Few-Shot Text Classification Reality Check
This paper compares all neural few-shot classification models, first adapting those made in the field of image processing to NLP, and second providing them access to transformers, and reveals that a simple baseline is surprisingly strong. Expand
ProtAugment: Unsupervised diverse short-texts paraphrasing for intent detection meta-learning
This work proposes PROTAUGMENT, a meta-learning algorithm for short texts classification applied to the intent detection task, a novel extension of Prototypical Networks that limits overfitting on the bias introduced by the few-shots classification objective at each episode. Expand