V0XNIHILI / few-shot-spoken-term-classification

A quick reimplementation of the two datasets ("digits" and "commands") proposed in the paper "An Investigation of Few-Shot Learning in Spoken Term Classification"

Few-shot spoken term classification dataset

A quick reimplementation of the two datasets ("digits" and "commands") proposed in the paper "An Investigation of Few-Shot Learning in Spoken Term Classification" by Yanbin Chen, Tom Ko, Lifeng Shang, Xiao Chen, Xin Jiang, and Qing Li: https://arxiv.org/abs/1812.10233. You can find their original code here.

In this version here, the data underneath the original test-train-val splits from the 12-way SpeechCommands task was adhered to for creating the meta-train, meta-val, and meta-test splits.

Installation

pip install git+https://github.com/V0XNIHILI/few-shot-spoken-term-classification

License

The license for this code alone is MIT.