2 stage 1-数据准备

if [ $stage -le 1 ]; then
  # format the data as Kaldi data directories
  for part in dev-clean-2 train-clean-5; do
    # use underscore-separated names in data directories.
    local/data_prep.sh $data/LibriSpeech/$part data/$(echo $part | sed s/-/_/g)
  done

  local/prepare_dict.sh --stage 3 --nj 30 --cmd "$train_cmd" \
    data/local/lm data/local/lm data/local/dict_nosp

  utils/prepare_lang.sh data/local/dict_nosp \
    "<UNK>" data/local/lang_tmp_nosp data/lang_nosp

  local/format_lms.sh --src-dir data/lang_nosp data/local/lm
  # Create ConstArpaLm format language model for full 3-gram and 4-gram LMs
  utils/build_const_arpa_lm.sh data/local/lm/lm_tglarge.arpa.gz \
    data/lang_nosp data/lang_nosp_test_tglarge
fi

源代码如上，主要的几个sh，分别以子文章的方式发出