Lhx94As / JSTSP_w2v_for_LID

Source code for: Efficient Self-supervised Learning Representations for Spoken Language Identification

JSTSP

Efficient Self-supervised Learning Representations for Spoken Language Identification.

Before running the training scripts,

1. Download asv-subtools and move the folder "subtool" to your kaldi/esg/
2. Download s3prl and run the pre_processing.py in s3prl to extract the features.

An example:

python pre_processing.py --json xsa_config.json
python train_xsa.py --json xsa_config.json

Before running, pls revise the json configuration file according to your own root.
Mainly the "Input" part, you can keep others since they are the parameters I am using :)