Speech Recognition
dataset: aishell1
Setup:
use docker: docker pull tongjinle/tasr:v1
mkdir lm
download[this file](https://deepspeech.bj.bcebos.com/zh_lm/zh_giga.no_cna_cmn.prune01244.klm)and put it into lm/
use:
download aishell:
bash scripts/data/download_aishell.sh
preprocess:
bash scripts/data/*
bash scripts/train/*
result: beam decode with lm cer:7.2
curves:
