Nearest Neighbour Few-Shot Learning for Cross-lingual Classification
This addresses the challenge of under-resourced languages in NLP by enabling more effective cross-lingual adaptation with minimal data, though it is incremental as it builds on existing few-shot and nearest neighbor methods.
The paper tackles the problem of over-fitting in cross-lingual classification when fine-tuning pre-trained models with few target samples, by proposing a nearest neighbor few-shot inference technique, which consistently improves performance over traditional fine-tuning using only a handful of labeled samples across 16 languages and two NLP tasks.
Even though large pre-trained multilingual models (e.g. mBERT, XLM-R) have led to significant performance gains on a wide range of cross-lingual NLP tasks, success on many downstream tasks still relies on the availability of sufficient annotated data. Traditional fine-tuning of pre-trained models using only a few target samples can cause over-fitting. This can be quite limiting as most languages in the world are under-resourced. In this work, we investigate cross-lingual adaptation using a simple nearest neighbor few-shot (<15 samples) inference technique for classification tasks. We experiment using a total of 16 distinct languages across two NLP tasks- XNLI and PAWS-X. Our approach consistently improves traditional fine-tuning using only a handful of labeled samples in target locales. We also demonstrate its generalization capability across tasks.