Paper, Code and Implementations
Few-shot word acquisition on SpokenCOCO
Before getting started, download the QbERT code for pair mining and extract within the model project folder. For example, after cloning the QbERT repo, your directory structure should look like this:
few-shot_word_acquisition
↳ 5-shot_5-way
↳ 5-shot_40-way
↳ 10-shot_5-way
↳ 50-shot_5-way
↳ 100-shot_5-way
↳ QbERT
To get the exact support sets we use, download them here and extract in the specific model folder. For example:
100-shot_5-way/support_sets
Download the specific models here:
Few-shot word acquisition for low-resource languages