Initial commit

asr.sh

+../../TEMPLATE/asr1/asr.sh
\ No newline at end of file

cmd.sh

+# ====== About run.pl, queue.pl, slurm.pl, and ssh.pl ======
+# Usage: <cmd>.pl [options] JOB=1:<nj> <log> <command...>
+# e.g.
+#   run.pl --mem 4G JOB=1:10 echo.JOB.log echo JOB
+#
+# Options:
+#   --time <time>: Limit the maximum time to execute.
+#   --mem <mem>: Limit the maximum memory usage.
+#   -–max-jobs-run <njob>: Limit the number parallel jobs. This is ignored for non-array jobs.
+#   --num-threads <ngpu>: Specify the number of CPU core.
+#   --gpu <ngpu>: Specify the number of GPU devices.
+#   --config: Change the configuration file from default.
+#
+# "JOB=1:10" is used for "array jobs" and it can control the number of parallel jobs.
+# The left string of "=", i.e. "JOB", is replaced by <N>(Nth job) in the command and the log file name,
+# e.g. "echo JOB" is changed to "echo 3" for the 3rd job and "echo 8" for 8th job respectively.
+# Note that the number must start with a positive number, so you can't use "JOB=0:10" for example.
+#
+# run.pl, queue.pl, slurm.pl, and ssh.pl have unified interface, not depending on its backend.
+# These options are mapping to specific options for each backend and
+# it is configured by "conf/queue.conf" and "conf/slurm.conf" by default.
+# If jobs failed, your configuration might be wrong for your environment.
+#
+#
+# The official documentaion for run.pl, queue.pl, slurm.pl, and ssh.pl:
+#   "Parallelization in Kaldi": http://kaldi-asr.org/doc/queue.html
+# =========================================================~
+
+
+# Select the backend used by run.sh from "local", "stdout", "sge", "slurm", or "ssh"
+cmd_backend='sge'
+
+# Local machine, without any Job scheduling system
+if [ "${cmd_backend}" = local ]; then
+
+    # The other usage
+    export train_cmd="run.pl"
+    # Used for "*_train.py": "--gpu" is appended optionally by run.sh
+    export cuda_cmd="run.pl"
+    # Used for "*_recog.py"
+    export decode_cmd="run.pl"
+
+# Local machine logging to stdout and log file, without any Job scheduling system
+elif [ "${cmd_backend}" = stdout ]; then
+
+    # The other usage
+    export train_cmd="stdout.pl"
+    # Used for "*_train.py": "--gpu" is appended optionally by run.sh
+    export cuda_cmd="stdout.pl"
+    # Used for "*_recog.py"
+    export decode_cmd="stdout.pl"
+
+
+# "qsub" (Sun Grid Engine, or derivation of it)
+elif [ "${cmd_backend}" = sge ]; then
+    # The default setting is written in conf/queue.conf.
+    # You must change "-q g.q" for the "queue" for your environment.
+    # To know the "queue" names, type "qhost -q"
+    # Note that to use "--gpu *", you have to setup "complex_value" for the system scheduler.
+
+    export train_cmd="queue.pl --mem 20G --config conf/queue.conf"
+    export cuda_cmd="queue.pl --mem 20G --config conf/queue.conf"
+    export decode_cmd="queue.pl"
+
+
+# "qsub" (Torque/PBS.)
+elif [ "${cmd_backend}" = pbs ]; then
+    # The default setting is written in conf/pbs.conf.
+
+    export train_cmd="pbs.pl"
+    export cuda_cmd="pbs.pl"
+    export decode_cmd="pbs.pl"
+
+
+# "sbatch" (Slurm)
+elif [ "${cmd_backend}" = slurm ]; then
+    # The default setting is written in conf/slurm.conf.
+    # You must change "-p cpu" and "-p gpu" for the "partition" for your environment.
+    # To know the "partion" names, type "sinfo".
+    # You can use "--gpu * " by defualt for slurm and it is interpreted as "--gres gpu:*"
+    # The devices are allocated exclusively using "${CUDA_VISIBLE_DEVICES}".
+
+    export train_cmd="slurm.pl"
+    export cuda_cmd="slurm.pl"
+    export decode_cmd="slurm.pl"
+
+elif [ "${cmd_backend}" = ssh ]; then
+    # You have to create ".queue/machines" to specify the host to execute jobs.
+    # e.g. .queue/machines
+    #   host1
+    #   host2
+    #   host3
+    # Assuming you can login them without any password, i.e. You have to set ssh keys.
+
+    export train_cmd="ssh.pl"
+    export cuda_cmd="ssh.pl"
+    export decode_cmd="ssh.pl"
+
+# This is an example of specifying several unique options in the JHU CLSP cluster setup.
+# Users can modify/add their own command options according to their cluster environments.
+elif [ "${cmd_backend}" = jhu ]; then
+
+    export train_cmd="queue.pl --mem 2G"
+    export cuda_cmd="queue-freegpu.pl --mem 2G --gpu 1 --config conf/queue.conf"
+    export decode_cmd="queue.pl --mem 4G"
+
+else
+    echo "$0: Error: Unknown cmd_backend=${cmd_backend}" 1>&2
+    return 1
+fi

conf/decode_asr.yaml

+lm_weight: 0.6
+ctc_weight: 0.4
+beam_size: 3
+penalty: 0.0
+maxlenratio: 0.0
+minlenratio: 0.0
+streaming: True

conf/decode_asr_streaming.yaml

+tuning/decode_asr_streaming.yaml
\ No newline at end of file

conf/fbank.conf

+--sample-frequency=16000 
+--num-mel-bins=80

conf/pbs.conf

+# Default configuration
+command qsub -V -v PATH -S /bin/bash
+option name=* -N $0
+option mem=* -l mem=$0
+option mem=0          # Do not add anything to qsub_opts
+option num_threads=* -l ncpus=$0
+option num_threads=1  # Do not add anything to qsub_opts
+option num_nodes=* -l nodes=$0:ppn=1
+default gpu=0
+option gpu=0
+option gpu=* -l ngpus=$0

conf/pitch.conf

+--sample-frequency=16000

conf/queue.conf

+# Default configuration
+command qsub -v PATH -cwd -S /bin/bash -j y -l arch=*64*
+option name=* -N $0
+option mem=* -l mem_free=$0,ram_free=$0
+option mem=0          # Do not add anything to qsub_opts
+option num_threads=* -pe smp $0
+option num_threads=1  # Do not add anything to qsub_opts
+option max_jobs_run=* -tc $0
+option num_nodes=* -pe mpi $0  # You must set this PE as allocation_rule=1
+default gpu=0
+option gpu=0
+option gpu=* -l gpu=$0 -q Default

conf/slurm.conf

+# Default configuration
+command sbatch --export=PATH
+option name=* --job-name $0
+option time=* --time $0
+option mem=* --mem-per-cpu $0
+option mem=0
+option num_threads=* --cpus-per-task $0
+option num_threads=1 --cpus-per-task 1
+option num_nodes=* --nodes $0
+default gpu=0
+option gpu=0 -p cpu
+option gpu=* -p gpu --gres=gpu:$0 -c $0  # Recommend allocating more CPU than, or equal to the number of GPU
+# note: the --max-jobs-run option is supported as a special case
+# by slurm.pl and you don't have to handle it in the config file.

conf/train_asr_streaming_transformer.yaml

+tuning/train_asr_streaming_transformer.yaml
\ No newline at end of file

conf/train_asr_transformer.yaml

+tuning/train_asr_transformer.yaml
\ No newline at end of file

conf/train_lm.yaml

+tuning/train_lm_adam.yaml
\ No newline at end of file

conf/train_lm_transformer.yaml

+tuning/train_lm_transformer.yaml
\ No newline at end of file

conf/tuning/decode_asr_streaming.yaml

+# lm_weight: 0.3
+beam_size: 10
+penalty: 0.0
+maxlenratio: 0.0
+minlenratio: 0.0
+ctc_weight: 0.3
+sim_chunk_length: 512
+disable_repetition_detection: true
+decoder_text_length_limit: 0
+encoded_feat_length_limit: 0

conf/tuning/train_asr_conformer.yaml

+batch_type: numel
+batch_bins: 140000000
+accum_grad: 1
+max_epoch: 60
+patience: none
+init: xavier_uniform
+val_scheduler_criterion:
+    - valid
+    - acc
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+encoder: conformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 12
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.1
+    input_layer: conv2d
+    normalize_before: true
+    macaron_style: true
+    pos_enc_layer_type: "rel_pos"
+    selfattention_layer_type: "rel_selfattn"
+    activation_type: "swish"
+    use_cnn_module:  true
+    cnn_module_kernel: 31
+
+decoder: transformer
+decoder_conf:
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.1
+    src_attention_dropout_rate: 0.1
+
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+optim: adam
+optim_conf:
+    lr: 0.0015
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 25000
+
+frontend_conf:
+  n_fft: 512
+  hop_length: 256
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2
\ No newline at end of file

conf/tuning/train_asr_streaming_conformer.yaml

+# network architecture
+# encoder related
+encoder: contextual_block_conformer    # contextual_block_conformer is the core of streaming conformer
+encoder_conf:
+    output_size: 256    # dimension of attention
+    attention_heads: 4
+    linear_units: 2048  # the number of units of position-wise feed forward
+    num_blocks: 12      # the number of encoder blocks
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.0
+    input_layer: conv2d # encoder architecture type
+    normalize_before: true
+    activation_type: swish
+    macaron_style: true
+    use_cnn_module: true
+    cnn_module_kernel: 15
+    block_size: 40      # streaming configuration
+    hop_size: 16        # streaming configuration
+    look_ahead: 16      # streaming configuration
+    init_average: true  # streaming configuration
+    ctx_pos_enc: true   # streaming configuration
+
+# decoder related
+decoder: transformer
+decoder_conf:
+    attention_heads: 4
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.0
+    src_attention_dropout_rate: 0.0
+
+# hybrid CTC/attention
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1     # label smoothing option
+    length_normalized_loss: false
+
+# minibatch related
+# batch_type: folded
+# batch_size: 128
+batch_type: numel
+batch_bins: 50000000
+#valid_batch_size: 1
+# optimization related
+accum_grad: 1
+grad_clip: 5
+patience: 3
+max_epoch: 50
+val_scheduler_criterion:
+    - valid
+    - acc
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+early_stopping_criterion:
+    - valid
+    - cer_ctc
+    - min
+keep_nbest_models: 10
+
+# NoamLR is deprecated. Use WarmupLR.
+# The following is equivalent setting for NoamLR:
+#
+#    optim: adam
+#    optim_conf:
+#        lr: 10.
+#    scheduler: noamlr
+#    scheduler_conf:
+#        model_size: 256
+#        warmup_steps: 25000
+#
+optim: adam
+optim_conf:
+    lr: 0.0005
+scheduler: warmuplr     # pytorch v1.1.0+ required
+scheduler_conf:
+    warmup_steps: 30000
+num_att_plot: 0
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2
\ No newline at end of file

conf/tuning/train_asr_streaming_transformer.yaml

+# This configuration requires 4 GPUs with 32GB memory
+batch_type: numel
+batch_bins: 140000000
+accum_grad: 6
+max_epoch: 33
+patience: none
+# The initialization method for model parameters
+init: xavier_uniform
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+encoder: contextual_block_transformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 18
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.1
+    input_layer: conv2d6
+    normalize_before: true
+    block_size: 40
+    hop_size: 16
+    look_ahead: 16
+    init_average: true
+    ctx_pos_enc: true
+
+decoder: transformer
+decoder_conf:
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.1
+    src_attention_dropout_rate: 0.1
+
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+optim: adam
+optim_conf:
+    lr: 0.002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 25000
+num_att_plot: 0
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2

conf/tuning/train_asr_transformer.yaml

+# encoder related
+encoder: transformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 12
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.0
+    input_layer: conv2d
+    normalize_before: true
+
+# decoder related
+decoder: transformer
+decoder_conf:
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.0
+    src_attention_dropout_rate: 0.0
+
+# hybrid CTC/attention
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1     # label smoothing option
+    length_normalized_loss: false
+
+# minibatch related
+# batch_type: folded
+# batch_size: 32
+
+# optimization related
+batch_type: numel
+batch_bins: 16000000
+accum_grad: 4
+# grad_clip : 5
+max_epoch: 200
+# patience: 3
+patience: none
+init: xavier_uniform
+val_scheduler_criterion:
+    - valid
+    - acc
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+# NoamLR is deprecated. Use WarmupLR.
+optim: adam
+optim_conf:
+    lr: 0.002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 25000
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2

conf/tuning/train_asr_transformer3.yaml

+batch_type: numel
+batch_bins: 15000000
+accum_grad: 6
+max_epoch: 100
+patience: none
+# The initialization method for model parameters
+init: xavier_uniform
+best_model_criterion:
+-   - valid
+    - acc
+    - max
+keep_nbest_models: 10
+
+encoder: transformer
+encoder_conf:
+    output_size: 512
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 18
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    attention_dropout_rate: 0.1
+    input_layer: conv2d6
+    normalize_before: true
+
+decoder: transformer
+decoder_conf:
+    attention_heads: 8
+    linear_units: 2048
+    num_blocks: 6
+    dropout_rate: 0.1
+    positional_dropout_rate: 0.1
+    self_attention_dropout_rate: 0.1
+    src_attention_dropout_rate: 0.1
+
+model_conf:
+    ctc_weight: 0.3
+    lsm_weight: 0.1
+    length_normalized_loss: false
+
+optim: adam
+optim_conf:
+    lr: 0.002
+scheduler: warmuplr
+scheduler_conf:
+    warmup_steps: 25000
+
+specaug: specaug
+specaug_conf:
+    apply_time_warp: true
+    time_warp_window: 5
+    time_warp_mode: bicubic
+    apply_freq_mask: true
+    freq_mask_width_range:
+    - 0
+    - 30
+    num_freq_mask: 2
+    apply_time_mask: true
+    time_mask_width_range:
+    - 0
+    - 40
+    num_time_mask: 2

conf/tuning/train_lm_adam.yaml

+lm_conf:
+    nlayers: 4
+    unit: 2048
+optim: adam
+optim_conf:
+    lr: 0.001
+batch_type: folded
+batch_size: 400   # batch size in LM training
+max_epoch: 20     # if the data size is large, we can reduce this
+patience: 3
+
+best_model_criterion:
+-   - valid
+    - loss
+    - min
+keep_nbest_models: 1

conf/tuning/train_lm_transformer.yaml

+# Trained with Nvidia TESLA V100, with 16GM RAM, x4
+lm: transformer
+lm_conf:
+    pos_enc: null
+    embed_unit: 128
+    att_unit: 512
+    head: 8
+    unit: 2048
+    layer: 16
+    dropout_rate: 0.1
+
+# optimization related
+grad_clip: 5.0
+batch_type: numel
+batch_bins: 2000000
+accum_grad: 1
+max_epoch: 15  # 15epoch is enougth
+
+optim: adam
+optim_conf:
+   lr: 0.001
+scheduler: warmuplr
+scheduler_conf:
+   warmup_steps: 25000
+
+best_model_criterion:
+-   - valid
+    - loss
+    - min
+keep_nbest_models: 10  # 10 is good.
\ No newline at end of file

db.sh

+../../TEMPLATE/asr1/db.sh
\ No newline at end of file

decode.sh

+../asr1/decode.sh
\ No newline at end of file

local/data.sh

+#!/bin/bash
+# Set bash to 'debug' mode, it will exit on :
+# -e 'error', -u 'undefined variable', -o ... 'error in pipeline', -x 'print commands',
+set -e
+set -u
+set -o pipefail
+
+log() {
+    local fname=${BASH_SOURCE[1]##*/}
+    echo -e "$(date '+%Y-%m-%dT%H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+SECONDS=0
+
+
+stage=0
+nj=32
+stop_stage=100000
+train_set=train_all
+valid_set=dev_all
+
+test_sets="librispeech-test_clean librispeech-test_other \
+        NER-Trs-Vol1-test NER-Trs-Vol2-test NER-Trs-Vol3-test NER-Trs-Vol4-test \
+        OC16-CE80 MATBN-test thchs30-test \
+        tat-vol1-test tat-vol2-test tat-edu-test"
+
+lid=false # whether to use language id as additional label
+use_noise=false
+global_path=
+nlsyms_txt=data/local/nlsyms.txt
+
+log "$0 $*"
+. ./path.sh || exit 1
+. ./cmd.sh || exit 1
+. utils/parse_options.sh || exit 1
+
+# if [ $# -ne 0 ]; then
+#     log "Error: No positional arguments are required."
+#     exit 2
+# fi
+
+if [ ${stage} -le 1 ]; then
+  # combine the sub sets into the train_set
+  echo "Stage 0: Combine Multiple Train Data Source"
+  utils/combine_data.sh --extra-files utt2num_frames data/${train_set} \
+      data/train-data/{OC16-CE80,TCC300,aishell-train} \
+      data/train-data/NER-Trs-Vol{1,2,3,4}-train \
+      data/train-data/{librispeech-train_100,librispeech-train_360} \
+      data/train-data/{tat-vol1-train,tat-vol2-train,tat-edu}
+  if $lid; then
+    mv data/${train_set}/text data/${train_set}/text.bak
+    python3 tools/add_lid_tag.py \
+      -utt data/train-data/tat-vol1-train/utt2spk \
+      -utt data/train-data/tat-vol2-train/utt2spk \
+      -utt data/train-data/tat-tmp/utt2spk \
+      -utt data/train-data/G2019429-fix/utt2spk \
+      -utt data/train-data/G2019432-fix/utt2spk \
+      -utt data/train-data/G2019459-fix/utt2spk \
+      -utt data/train-data/G2019463-fix/utt2spk \
+      -utt data/train-data/G2019479-fix/utt2spk \
+      data/${train_set}/text.bak tools/taiwanese.v2.csv data/${train_set}/text
+    rm data/${train_set}/text.bak
+    utils/fix_data_dir.sh data/${train_set}
+  fi
+
+  # combine the sub sets into the dev_set
+  echo "Stage 0: Combine Multiple Dev Data Source"
+  utils/combine_data.sh --extra-files utt2num_frames data/${valid_set} \
+      data/train-data/NER-Trs-Vol{1,2,3,4}-eval \
+      data/train-data/{thchs30-dev,aishell-dev} \
+      data/train-data/{librispeech-dev_clean,librispeech-dev_other} \
+      data/train-data/{tat-vol1-dev,tat-vol2-dev}
+      # data/train-data/{NER-Trs-Vol1-eval,librispeech-dev_clean,tat-vol1-dev}
+  
+  if $lid; then
+    mv data/${valid_set}/text data/${valid_set}/text.bak
+    python3 tools/add_lid_tag.py \
+      -utt data/train-data/tat-vol1-dev/utt2spk \
+      -utt data/train-data/tat-vol2-dev/utt2spk \
+      data/${valid_set}/text.bak tools/taiwanese.v2.csv data/${valid_set}/text
+    rm data/${valid_set}/text.bak
+    utils/fix_data_dir.sh data/${valid_set}
+  fi
+
+  if $use_noise; then
+    echo "Use FaNT to add noise"
+    train_set=${train_set}_noise
+    rirsdir=
+    noise_opt=()
+    dir_opt=()
+    dest_opt=()
+
+    # use FaNT to increase data diversity
+    noise_opt+=("/nfs/TS-1635AX/Corpora/musan")
+    noise_opt+=("/nfs/TS-1635AX/Corpora/NOISE_DATASETs/TRAIN")
+    for n in ${noise_opt[@]}; do
+      srcdir=data/train_all
+      _n=$(echo $n | awk -F'/' '{print $NF}')
+      local/multi_condition/perturb_data_dir_fant_convert.sh --nj $nj \
+                                                            --noisedir $n \
+                                                            $srcdir
+      dest_opt+=(${srcdir}_fant_${_n})
+    done
+    utils/data/combine_data.sh data/train_all_fant ${dest_opt[@]}
+    rm -r ${dest_opt[@]}
+
+    dir_opt+=(data/train_all_fant)
+    dir_opt+=(data/train_all)
+
+    # add RIRs, simulated RIRs, isotropic noises and point-source noises
+    if [[ -n "$rirsdir" ]]; then
+      srcdir=data/train_all_fant
+      samplerate=16000
+      # Make a version with reverberated speech
+      rvb_opts=()
+      rvb_opts+=(--rir-set-parameters "0.5, ${rirsdir}/simulated_rirs/smallroom/rir_list")
+      rvb_opts+=(--rir-set-parameters "0.5, ${rirsdir}/simulated_rirs/mediumroom/rir_list")
+      # Make a reverberated version of the SWBD+SRE list. Note that we don't add any
+      # additive noise here.
+      python3 steps/data/reverberate_data_dir.py "${rvb_opts[@]}" \
+                                                --prefix "reverb" \
+                                                --speech-rvb-probability 1 \
+                                                --pointsource-noise-addition-probability 0 \
+                                                --isotropic-noise-addition-probability 0 \
+                                                --num-replications 1 \
+                                                --source-sampling-rate $samplerate \
+                                                ${srcdir} ${srcdir}_reverb
+      dir_opt+=(${srcdir}_reverb)
+    fi
+    utils/data/combine_data.sh data/${train_set} ${dir_opt[@]}
+    utils/fix_data_dir.sh data/${train_set}
+  fi
+fi
+
+if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ] && $lid; then
+    log "stage 2: Create Non-linguistic Symbols for Language ID"
+    cut -f 2- data/${train_set}/text | grep -o -P '\[.*?\]|\<.*?\>' | sort | uniq > ${nlsyms_txt}
+    log "save non-linguistic symbols in ${nlsyms_txt}"
+fi
+
+#if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
+#  # use external data
+#  echo "$0: preparing extra corpus for subword LM training..."
+#  mkdir -p data/local/other_text
+#  local/lm/prepare_extra_text.sh --normjobs $nj \
+#                              --global-path $global_path \
+#                              data/local/lm data/local/lm/corpus || exit 1
+#  if [ ! -e data/local/other_text/text ]; then
+#    # provide utterance id to each texts
+#    find data/local/lm/norm -mindepth 1 -maxdepth 3 -type f | xargs cat > data/local/other_text/text
+#  fi
+#fi
+
+if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
+  for test in ${test_sets}; do
+    cp -r data/test-data/${test} data
+    utils/fix_data_dir.sh data/${test}
+
+    if $lid; then
+      echo "Stage 4: add language ID to Dev Data Source"
+      mv data/${test}/text data/${test}/text.bak
+      python3 tools/add_lid_tag.py \
+        -utt data/test-data/tat-vol1-test/utt2spk \
+        -utt data/test-data/tat-vol2-test/utt2spk \
+        data/${test}/text.bak tools/taiwanese.csv data/${test}/text
+      rm data/${test}/text.bak
+      utils/fix_data_dir.sh data/${test}
+    fi
+  done
+fi
+
+
+
+log "Successfully finished. [elapsed=${SECONDS}s]"

local/path.sh

+# We only need the NCTU parser stuff below for the optional text normalization(for LM-training) step
+PARSER_ROOT=tools/new_parser_UNICODE
+export PATH=$PATH:$PARSER_ROOT
+
+# FaNT is needed for noise/phone
+FANT_ROOT=tools/fant
+export PATH=$PATH:$FANT_ROOT
+
+# g729a/b is needed for noise/phone
+G729_ROOT=tools/g729a
+export PATH=$PATH:$G729_ROOT

path.sh

+../../TEMPLATE/asr1/path.sh
\ No newline at end of file

pyscripts

+../../TEMPLATE/asr1/pyscripts
\ No newline at end of file

run.sh

+#!/bin/bash
+# Set bash to 'debug' mode, it will exit on :
+# -e 'error', -u 'undefined variable', -o ... 'error in pipeline', -x 'print commands',
+set -e
+set -u
+set -o pipefail
+
+# E2E model related
+train_set=train_all
+valid_set=dev_all
+test_sets="librispeech-test_clean librispeech-test_other \
+        tat-vol1-test tat-vol2-test tat-tmp-edu \
+        NER-Trs-Vol1-test NER-Trs-Vol2-test NER-Trs-Vol3-test NER-Trs-Vol4-test"
+use_noise=false
+global_path=`pwd`
+lid=true # whether to use language id as additional label
+nj=150
+stage=0
+stop_stage=10000
+
+. ./path.sh
+. ./cmd.sh
+. ./utils/parse_options.sh
+
+asr_config=conf/tuning/train_asr_conformer.yaml
+lm_config=conf/tuning/train_lm_transformer.yaml
+inference_config=conf/decode_asr.yaml
+nlsyms_txt=data/local/nlsyms.txt
+
+if "${use_noise}"; then train_set=${train_set}_noise; fi
+#--use_streaming true \
+./asr.sh \
+    --stage $stage \
+    --stop_stage $stop_stage \
+    --use_lm false \
+    --nj $nj \
+    --lang cht_eng_tw \
+    --ngpu 10 \
+    --num_nodes 1 \
+    --nbpe 20000 \
+    --token_type bpe \
+    --feats_type raw \
+    --audio_format wav \
+    --max_wav_duration 30 \
+    --speed_perturb_factors "0.9 1.0 1.1" \
+    --asr_config "${asr_config}" \
+    --lm_config "${lm_config}" \
+    --inference_config "${inference_config}" \
+    --local_data_opts "--global-path $global_path --nj $nj --nlsyms_txt ${nlsyms_txt} --lid ${lid}" \
+    --train_set "${train_set}" \
+    --valid_set "${valid_set}" \
+    --test_sets "${test_sets}" \
+    --bpe_nlsyms "[CHT],[EN],[TW]" \
+    --lm_train_text "data/${train_set}/text" \
+    --bpe_train_text "data/${train_set}/text" "$@" \
+    --local_score_opts "--score_lang_id ${lid}" "$@"
+    

run_decode.sh

+#!/bin/bash
+# Set bash to 'debug' mode, it will exit on :
+# -e 'error', -u 'undefined variable', -o ... 'error in pipeline', -x 'print commands',
+set -e
+set -u
+set -o pipefail
+
+# E2E model related
+train_set=train_all
+valid_set=dev_all
+test_sets="librispeech-test_clean librispeech-test_other \
+        NER-Trs-Vol1-test NER-Trs-Vol2-test NER-Trs-Vol3-test \
+        NER-Trs-Vol4-test OC16-CE80 MATBN-test thchs30-test"
+use_noise=false
+global_path=`pwd`
+lid=true # whether to use language id as additional label
+nj=100
+
+. ./path.sh
+. ./cmd.sh
+. ./utils/parse_options.sh
+
+asr_config=conf/tuning/train_asr_streaming_conformer.yaml
+lm_config=conf/tuning/train_lm_transformer.yaml
+inference_config=conf/decode_asr_streaming.yaml
+nlsyms_txt=data/local/nlsyms.txt
+
+if "${use_noise}"; then train_set=${train_set}_noise; fi
+
+./decode.sh \
+    --stage 2 \
+    --stop_stage 10000 \
+    --use_streaming true \
+    --gpu_inference false \
+    --inference_nj $nj \
+    --use_lm false \
+    --nj $nj \
+    --lang cht_eng_tw.v2 \
+    --ngpu 10 \
+    --num_nodes 1 \
+    --nbpe 30000 \
+    --token_type bpe \
+    --feats_type raw \
+    --audio_format wav \
+    --speed_perturb_factors "0.9 1.0 1.1" \
+    --asr_config "${asr_config}" \
+    --lm_config "${lm_config}" \
+    --inference_config "${inference_config}" \
+    --local_data_opts "--global-path $global_path --stage 1" \
+    --train_set "${train_set}" \
+    --valid_set "${valid_set}" \
+    --test_sets "${test_sets}" \
+    --asr_speech_fold_length 512 \
+    --asr_text_fold_length 150 \
+    --lm_fold_length 150 \
+    --bpe_nlsyms "[CHT],[EN],[TW]" \
+    --lm_train_text "data/${train_set}/text" \
+    --bpe_train_text "data/${train_set}/text" "$@" \
+    --local_score_opts "--score_lang_id ${lid}" "$@"
+    

scripts

+../../TEMPLATE/asr1/scripts
\ No newline at end of file

steps

+../../../tools/kaldi/egs/wsj/s5/steps
\ No newline at end of file

tools/add_lid_tag.py

+import os, sys
+import argparse
+import csv
+import re
+
+def parse_opts():
+    parser = argparse.ArgumentParser(
+        description='Strips unhelpful, from LM viewpoint, strings from PG texts',
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter)
+    parser.add_argument('-utt', '--taiwanese-utt-file', action='append',
+                        help='Split chinese word to char which use for writting the output text')
+
+    parser.add_argument('in_text', type=str, help='Input text file')
+    parser.add_argument('taiwanese_tab', type=str, help='Input taiwanese table file')
+    parser.add_argument('out_text', type=str, help='Filtered output text file')
+    opts = parser.parse_args()
+    return opts
+
+def check_english(check_str):
+    check = False
+    for ch in check_str:
+        if (ch >= u'\u0041' and ch <= u'\u005A') or \
+            (ch >= u'\u0061' and ch <= u'\u007A') or ch == "'":
+            check = True
+        else:
+            return False
+    if check:
+        return True
+
+def check_full_english(check_str):
+    check = False
+    for w in check_str.split():
+        if check_english(w):
+            check = True
+        else:
+            return False
+    if check:
+        return True
+
+def check_contain_chinese(check_str):
+    check = False
+    RE_HANS = re.compile(
+        r'^(?:['
+        r'\u3100-\u312f'           # Bopomofo
+        r'\u3400-\u4dbf'           # CJK Ext.A:[3400-4DBF]
+        r'\u4e00-\u9fff'           # CJK baise:[4E00-9FFF]
+        r'\uf900-\ufaff'           # CJK Comp:[F900-FAFF]
+        r'\U00020000-\U0002A6DF'   # CJK Ext.B:[20000-2A6DF]
+        r'\U0002A703-\U0002B73F'   # CJK Ext.C:[2A700-2B73F]
+        r'\U0002B740-\U0002B81D'   # CJK Ext.D:[2B740-2B81D]
+        r'\U0002F80A-\U0002FA1F'   # CJK Comp:[2F800-2FA1F]
+        r'])+$'
+    )
+    for ch in check_str:
+        if RE_HANS.match(ch):
+            check = True
+        else:
+            return False
+    if check:
+        return True
+
+def read_csv(path):
+    table = []
+    with open(path, newline='') as csvfile:
+        rows = csv.reader(csvfile)
+        rows = list(rows)
+        con = rows[0][2:]
+        table.extend(con)
+        for row in rows[2:]:
+            table.extend([ row[0]+_n for _n in con ])
+    return table
+
+def check_taiwanese(check_str, table):
+    tone_list = [u"ā", u"á", u"ǎ", u"à", u"â", u"a̍", \
+                 u"ē", u"é", u"ě", u"è", u"ê", u"e̍", \
+                 u"ō", u"ó", u"ǒ", u"ò", u"ô", u"o̍", u"ő", \
+                 u"ī", u"í", u"ǐ", u"ì", u"î", u"i̍", \
+                 u"ū", u"ú", u"ǔ", u"ù", u"û", u"u̍", \
+                       u"ń", u"ň", u"ǹ", u"n̍", \
+                 u"m̄", u"ḿ", u"m̀", u"m̍"]
+    # tone_list = [u"ā", u"á", u"ǎ", u"à", u"â", \
+    #              u"ē", u"é", u"ě", u"è", \
+    #              u"ō", u"ó", u"ǒ", u"ò", u"ô", u"ő", \
+    #              u"ī", u"í", u"ǐ", u"ì", u"i", u"î", \
+    #              u"ū", u"ú", u"ǔ", u"ù", u"ü", u"ǖ", u"ǘ", u"ǚ" ,u"ǜ", u"û", \
+    #              u"ń", u"ň", u"ǹ", \
+    #              u"m̄", u"ḿ", u"m̀", \
+    #              u"ê", u"ê̄", u"ế", u"ê̌", u"ề"]
+    if len(check_str.split()) == 2:
+        check = False
+        for _str in check_str.split():
+            if not ( any( [ _ in tone_list for _ in _str] ) or _str in table or check_contain_chinese(_str) ):
+                return False
+        return True
+    else:
+        return any([ _ in tone_list for _ in check_str] ) or check_str in table
+
+if __name__ == '__main__':
+    opts = parse_opts()
+
+    tai_id = []
+    if opts.taiwanese_utt_file:
+        for f in opts.taiwanese_utt_file:
+            with open(f, "r") as rf:
+                tai_id.extend([ l.split()[0] for l in rf.readlines() ])
+
+    data = []
+    table = read_csv(opts.taiwanese_tab)
+    with open(opts.in_text, "r", encoding="utf-8") as rf:
+        for l in rf.readlines():
+            if len(l.split()) > 1:
+                u = l.split(maxsplit=1)[0]
+
+                # taiwanese mode
+                taiwanese_mode = False
+                if opts.taiwanese_utt_file:
+                    if u in tai_id:
+                        taiwanese_mode = True
+
+                t = ""
+                switchcode = ""
+                b_w = ""
+                try:
+                    for w in l.split(maxsplit=1)[1].split():
+                        if len(t) == 0:
+                            if taiwanese_mode:
+                                if check_taiwanese(w, table) or check_contain_chinese(w):
+                                    t += "[TW] @{} ".format(w)
+                                    switchcode = "tw"
+                                else:
+                                    t += "[EN] {} ".format(w)
+                                    switchcode = "en"
+                            else:
+                                if check_contain_chinese(w):
+                                    t += "[CHT] {} ".format(w)
+                                    switchcode = "cht"
+                                else:
+                                    t += "[EN] {} ".format(w)
+                                    switchcode = "en"
+                        else:
+                            if taiwanese_mode:
+                                if switchcode == "tw":
+                                    if check_taiwanese("{} {}".format(b_w, w), table):
+                                        t += " @{} ".format(w)
+                                    else:
+                                        t += "[EN] {} ".format(w)
+                                        switchcode = "en"
+                                else:
+                                    if check_full_english(b_w+w) and not check_taiwanese("{} {}".format(b_w, w), table):
+                                        t += " {} ".format(w)
+                                    else:
+                                        t += "[TW] @{} ".format(w)
+                                        switchcode = "tw"
+                            else:
+                                if switchcode == "cht":
+                                    if check_contain_chinese(b_w+w):
+                                        t += " {} ".format(w)
+                                    else:
+                                        t += "[EN] {} ".format(w)
+                                        switchcode = "en"
+                                else:
+                                    if check_full_english(b_w+w):
+                                        t += " {} ".format(w)
+                                    else:
+                                        t += "[CHT] {} ".format(w)
+                                        switchcode = "cht"
+                            #print(b_w+w, switchcode)
+                        b_w = w
+                    data.append("{} {}\n".format(u, " ".join(_t for _t in t.split()).strip()))
+                except Exception as e:
+                    print(u, e)
+
+    with open(opts.out_text, "w", encoding="utf-8") as wf:
+        wf.writelines(data)

tools/taiwanese.csv

+INDEX,NONE,a,ah,ai,aih,ainn,ak,am,an,ang,ann,annh,ap,at,au,auh,aunnh,e,eh,enn,ennh,i,ia,iah,iak,iam,ian,iang,iann,iannh,iap,iat,iau,iauh,ih,ik,im,in,ing,inn,io,ioh,iok,iong,ip,it,iu,iuh,iunn,m,mh,ng,ngh,o,oh,ok,om,ong,onn,onnh,oo,ooh,op,u,ua,uah,uai,uainn,uan,uang,uann,uat,ue,ueh,uh,ui,uinn,un,ut
+NONE,,a,ah,ai,aih,ainn,ak,am,an,ang,ann,annh,ap,at,au,auh,aunnh,E,eh,enn,ennh,i,ia,iah,iak,iam,ian,iang,iann,iannh,iap,iat,iau,iauh,ih,ik,im,ien,ieng,inn,io,ioh,iok,iong,ip,it,iu,iuh,iunn,nm,mh,eng,ngh,e,oh,ok,om,ong,onn,onnh,o,ooh,op,u,ua,uah,uai,uainn,uan,uang,uann,uat,oe,ueh,uh,ui,uinn,uen,ut
+p,b,b a,b ah,b ai,,,b ak,,b an,b ang,,,,b at,b au,,,b E,b eh,b enn,,b i,,b iah,b iak,,b ian,b iang,b iann,,,b iat,b iau,,b ih,b ik,,b ien,b ieng,b inn,b io,,,,,b it,b iu,,,,,b eng,,b e,b oh,b ok,,b ong,,,b o,,,b u,b ua,b uah,,,b uan,,b uann,b uat,b oe,b ush,b uh,b ui,,b uen,b ut
+ph,p,p a,p ah,p ai,,p ainn,p ak,,p an,p ang,p ann,,,,p au,p auh,,p E,,p enn,,p i,,p iah,p iak,,p ian,p iang,p iann,,,p iat,p iau,,p ih,p ik,,p ien,p ieng,p inn,p io,,,,,p it,b iu,,,,,,p ngh,p e,p oh,p ok,,p ong,,,p o,,,p u,p ua,p uah,,,p uan,,p uann,p uat,p oe,p ueh,p uh,p ui,,p uen,p ut
+m,m,m a,m ah,m ai,,,,,,,,,,,m au,m auh,,m E,m eh,,,m i,m ia,,,,,,,,,,m iau,,m ih,,,,,,,,,,,,,,,,,m eng,,,,,,,,,m o,m ooh,,,m ua,,m ue,,,,,,m oe,,,m ui,,,
+b,bh,bh a,bh ah,bh ai,,,bh ak,,bh an,bh ang,,,,bh at,bh au,,,bh E,bh eh,,,bh i,,,,,bh ian,,,,,bh iat,bh iau,,bh ih,b ik,,bh ien,bh ieng,,bh io,,,,,bh it,bh iu,,,,,,,bh e,,bh ok,,bh ong,,,bh o,,,bh u,bh ua,bh uah,,,bh uan,,,bh uat,bh oe,bh ueh,,bh ui,,bh uen,bh ut
+t,d,d a,d ah,d ai,,d ainn,d ak,d am,d an,d ang,d ann,,d ap,d at,d au,d auh,,d E,d eh,d enn,,d i,d ia,d iah,d iak,d iam,d ian,,d iann,,d iap,d iat,d iau,,d ih,d ik,d im,d ien,d ieng,d inn,d io,d ioh,d iok,d iong,,d it,diu,d iuh,d iunn,,,d eng,,d e,d oh,d ok,d om,d ong,,,d o,,,d u,d ua,d uah,,,d uan,,d uann,d uat,d oe,,d uh,d ui,,d uen,d ut
+th,t,t a,t ah,t ai,,,t ak,t am,t an,t ang,t ann,,t ap,t at,t au,,,t E,t eh,t enn,,t i,,t iah,,t iam,t ian,,t iann,,t iap,t iat,t iau,,t ih,t ik,t im,t ien,t ieng,t inn,t io,,t iok,t iong,,tsh it,t iu,,,,,t eng,,t e,t oh,t ok,,t ong,,,t o,,,t u,t ua,t uah,,,t uan,,t uann,t uat,th E,,t uh,t ui,,t uen,t ut
+n,n,n a,n ah,n ai,n ah,,,,,,,,,,n au,n auh,,n E,n eh,,,n i,n ia,,,,,,,,,,n iau,,n ih,,,,,,,,,,,,n iu,,,,,n eng,,ne,,,,,,,n o,,,,n ua,,,,,,,,,,,n ui,,,
+l,l,l a,l ah,l ai,,,l ak,l am,l an,l ang,,,l ap,l at,l au,l auh,,l E,l eh,,,l i,,l iah,l iok,l iam,l ian,l iang,,,l iap,l iat,l iau,,l ih,l ik,l im,l ien,l ieng,,l io,l ioh,l iok,l iong,l ip,l it,l iu,,,,,,,l e,l oh,l ok,,l ong,,,l o,,l op,l u,l ua,l uah,,,l uan,,,l uat,l oe,,l uh,l ui,,l uen,l ut
+k,g,g a,g ah,g ai,,g ainn,g ak,k am,g an,g ang,g ann,,g ap,g at,g au,g auh,,g E,g eh,g enn,,g i,g ia,g iah,,g iam,g ian,k iong,g iann,,g iap,g iat,g iua,,g ih,g ik,g im,g ien,g ieng,g inn,g io,g ioh,g iok,g iong,g ip,,g iu,,g iunn,,,g eng,,g e,g oh,g ok,,g ong,g onn,,g o,,,g u,g ua,g uah,g uai,g uainn,g uan,,g uann,g uat,g oe,g ueh,,g ui,,g uen,g ut
+kh,k,k a,k ah,k ai,,k ainn,k ak,k am,k an,k ang,k ann,,k ap,k at,k au,,,k E,k eh,k enn,k ennh,k i,k ia,k iah,k iak,k iam,k ian,k iang,,,k iap,k iat,k iau,,k ih,k ik,k im,k ien,k ieg,k inn,k io,k ioh,k iok,k iong,k ip,k it,k iu,,k iunn,,,k eng,k ngh,k e,,k ok,,k ong,,,k o,,,k u,k ua,k uah,k uai,,k uan,,k uann,k uat,k oe,k ueh,k uh,k ui,k uinn,k uen,k ut
+ng,ng,ng a,,ng ai,,,,,,,,,,,ng au,,,ng E,ng eh,,,,ng ia,,,,,,,,,,ng iau,ng iauh,,,,,,,,,,,,,ng iu,,,,,,,,,,,,,,ng o,,,,,,,,,,,,,,,,,,
+h,h,h a,h ah,h ai,h aih,h ainn,h ak,h am,h an,h ang,h ann,h annh,h ap,h at,h au,,h aunnh,h E,h eh,,,h i,h ia,h iah,,h iam,h ian,h iang,h iann,h iannh,h iap,h iat,h iau,h iauh,,h ik,h im,h ien,h ieg,h inn,h io,h ioh,h iok,h iong,h ip,h it,h iu,,h iunn,h nm,h mh,h eng,h ngh,h e,h oh,h ok,,h ong,h onn,h onnh,h o,h ooh,,h u,h ua,h uah,h uai,h uainn,h uan,,h uann,h uat,h oe,h ugh,h uh,h ui,,h uen,h ut
+g,gh,gh a,,gh ai,,,gh ak,gh am,gh an,gh ang,,,,,gh au,,,gh E,g eh,,,gh i,gh ia,gh iah,,gh iam,gh ian,gh iang,,,gh iap,gh iat,gh iau,,,gh ik,gh im,gh ien,gh ieng,,gh io,gh ioh,gh iok,gh iong,,,gh iu,,,,,,,gh e,,gh ok,,gh ong,,,gh o,,,gh u,gh ua,,,,gh uan,,,gh uat,gh oe,gh ueh,,gh ui,,gh ien,
+ts,ts,ts a,ts ah,ts ai,,ts ainn,ts ak,ts am,ts an,ts ang,ts ann,,ts ap,ts at,ts au,,,ts E,ts eh,ts enn,,ts i,ts ia,ts iah,,ts iam,ts ian,ts iang,ts iann,,ts iap,ts iat,ts iau,,ts ih,ts ik,ts im,ts ien,ts ieng,ts inn,ts io,ts ioh,ts iok,ts iong,ts ip,ts it,ts iu,ts iuh,ts iunn,,,ts eng,,ts e,ts oh,ts ok,,ts ong,,,ts o,,,ts u,ts ua,ts uah,,ts uainn,ts uan,,ts uann,ts uat,ts oe,ts ueh,ts uh,ts ui,,ts uen,ts ut
+tsh,tsh,tsh a,tsh ah,tsh ai,,,tsh ak,tsh am,tsh an,tsh ang,tsh ann,,tsh ap,tsh at,tsh au,tsh auh,,tsh E,tsh eh,tsh enn,,tsh i,tsh ia,tsh iah,tsh iak,tsh iam,tsh ian,tsh iang,tsh iann,,tsh iap,tsh iat,tsh iau,,tsh ih,tsh ik,tsh im,tsh ien,tsh ieng,tsh inn,tsh io,tsh ioh,tsh iok,tsh iong,x ip,tsh it,tsh iu,,tsh iunn,,,tsh eng,,tsh e,tsh oh,tsh ok,,tsh ong,,,tsh o,,,tsh u,tsh ua,tsh uah,,,tsh uan,tsh uang,tsh uann,,tsh oe,,tsh uh,tsh ui,,tsh uen,tsh ut
+s,s,s a,s ah,s ai,,,s ak,s ma,s an,s ang,s ann,s annh,s ap,s at,s au,,,s E,s eh,s enn,,s i,s ia,s iah,s iak,s iam,s ian,s iang,s iann,s iann,s iap,s iat,s iau,,s ih,s ik,s im,s ien,s ieng,s inn,s io,s ioh,s iok,s iong,s ip,s it,s iu,,s iunn,,,s eng,,s e,s oh,s ok,s om,s ong,,,s o,,,s u,s ua,s uah,,s uainn,s uan,,s uann,s uat,s oe,s ueh,s uh,s ui,,s uen,s ut
+j,nj,,,,,,,,,,,,,,,,,,,,,j i,j ia,j iah, ,j iam,j ian,j dang,,,j iap,j iat,j iau,,tsh ih,,j im,j ien,,,j io,,j iok,j iong,j ip,j it,j iu,,,,,,,,,,,,,,,,,j u,,j uah,,,,,,,j oe,,,,,j uen,

tools/taiwanese.v2.csv

+INDEX,NONE,a,ah,ai,aih,ainn,ak,am,an,ang,ann,annh,ap,at,au,auh,aunnh,e,eh,enn,ennh,i,ia,iah,iak,iam,ian,iang,iann,iannh,iap,iat,iau,iauh,ih,ik,im,in,ing,inn,io,ioh,iok,iong,ionn,ip,it,iu,iuh,iunn,m,mh,ng,ngh,o,oh,ok,om,ong,onn,oo,ooh,op,u,ua,uah,uai,uainn,uan,uann,uat,ue,ueh,uh,ui,uinn,un,ut
+NONE,,a,ah,ai,aih,ainn,ak,am,an,ang,ann,annh,ap,at,au,auh,aunnh,e,eh,enn,ennh,i,ia,iah,iak,iam,ian,iang,iann,iannh,iap,iat,iau,iauh,ih,ik,im,in,ing,inn,io,ioh,iok,iong,ionn,ip,it,iu,iuh,iunn,m,mh,ng,ngh,o,oh,ok,om,ong,onn,oo,ooh,op,u,ua,uah,uai,uainn,uan,uann,uat,ue,ueh,uh,ui,kh uinn,un,ut
+p,p,p a,p ah,p ai,,,p ak,,p an,p ang,,,,p at,p au,,,p e,p eh,p enn,,p i,,p iah,p iak,,p ian,p iang,p iann,,,p iat,p iau,,p ih,p ik,,p in,p ing,p inn,p io,,,,,,p it,p iu,,,,,p ng,,p o,p oh,p ok,,p ong,,p oo,,,p u,p ua,p uah,,,p uan,p uann,p uat,p ue,p ueh,p uh,p ui,,p un,p ut
+ph,ph,ph a,ph ah,ph ai,,ph ainn,ph ak,,ph an,ph ang,ph ann,,,,ph au,ph auh,,ph e,,ph enn,,ph i,,ph iah,ph iak,,ph ian,ph iang,ph iann,,,ph iat,ph iau,,ph ih,ph ik,,ph in,ph ing,ph inn,ph io,,,,,,ph it,,,,,,,ph ngh,ph o,ph oh,ph ok,,ph ong,,ph oo,,,ph u,ph ua,ph uah,,,ph uan,ph uann,ph uat,ph ue,ph ueh,ph uh,ph ui,,ph un,ph ut
+b,b,b a,b ah,b ai,,,b ak,,b an,b ang,,,,b at,b au,,,b e,b eh,,,b i,,,,,b ian,,,,,b iat,b iau,,b ih,b ik,,b in,b ing,,b io,,,,,,b it,b iu,,,,,,,b o,,b ok,,b ong,,b oo,,,b u,b ua,b uah,,,b uan,,b uat,b ue,b ueh,,b ui,,b un,b ut
+m,m,m a,m ah,m ai,,,,,,,,,,,m au,m auh,,m e,m eh,,,m i,m ia,,,,,,,,,,m iau,,m ih,,,,,,,,,,,,,,,,,,m ng,,,,,,,,m oo,m ooh,,,m ua,,,,,,,m ue,,,m ui,,,
+t,t,t a,t ah,t ai,,t ainn,t ak,t am,t an,t ang,t ann,,t ap,t at,t au,t auh,,t e,t eh,t enn,,t i,t ia,t iah,t iak,t iam,t ian,,t iann,,t iap,t iat,t iau,,t ih,t ik,t im,t in,t ing,t inn,t io,t ioh,t iok,t iong,t iunn,,t it,t iu,t iuh,t iunn,,,t ng,,t o,t oh,t ok,t om,t ong,,t oo,,,t u,t ua,t uah,,,t uan,t uann,t uat,t ue,,t uh,t ui,,t un,t ut
+th,th,th a,th ah,th ai,,,th ak,th am,th an,th ang,th ann,,th ap,th at,th au,,,th e,th eh,th enn,,th i,,th iah,,th iam,th ian,,th iann,,th iap,th iat,th iau,,th ih,th ik,th im,th in,th ing,th inn,th io,,th iok,th iong,,,,th iu,,,,,th ng,,th o,th oh,th ok,,th ong,,th oo,,,th u,th ua,th uah,,,th uan,th uann,th uat,,,th uh,th ui,,th un,th ut
+n,n,n a,n ah,n ai,,,,,,,,,,,n au,n auh,,n e,n eh,,,n i,n ia,,,,,,,,,,n iau,,n ih,,,,,,,,,,,,,n iu,,,,,n ng,,,,,,,,n oo,,,,n ua,,,,,,,,,,,,,
+l,l,l a,l ah,l ai,,,l ak,l am,l an,l ang,,,l ap,l at,l au,l auh,,l e,l eh,,,l i,,l iah,,l iam,l ian,l iang,,,l iap,l iat,l iau,,l ih,l ik,l im,l in,l ing,,l io,l ioh,l iok,l iong,,l ip,l it,l iu,,,,,,,l o,l oh,l ok,,l ong,,l oo,l ooh,l op,l u,l ua,l uah,,,l uan,,l uat,l ue,,l uh,l ui,,l un,l ut
+k,k,k a,k ah,k ai,,k ainn,k ak,k am,k an,k ang,k ann,,k ap,k at,k au,k auh,,k e,k eh,k enn,,k i,k ia,k iah,,k iam,k ian,,k iann,,k iap,k iat,k iau,,k ih,k ik,k im,k in,k ing,k inn,k io,k ioh,k iok,k iong,,k ip,,k iu,,k iunn,,,k ng,,k o,k oh,k ok,,k ong,k onn,k oo,,,k u,k ua,k uah,k uai,k uainn,k uan,k uann,k uat,k ue,k ueh,,k ui,,k un,k ut
+kh,kh,kh a,kh ah,kh ai,,kh ainn,kh ak,kh am,kh an,kh ang,kh ann,,kh ap,kh at,kh au,,kh aunnh,kh e,kh eh,kh enn,kh ennh,kh i,kh ia,kh iah,kh iak,kh iam,kh ian,kh iang,,,kh iap,kh iat,kh iau,,kh ih,kh ik,kh im,kh in,kh ing,kh inn,kh io,kh ioh,kh iok,kh iong,,kh ip,kh it,kh iu,,kh iunn,,,kh ng,kh ngh,kh o,,kh ok,,kh ong,,kh oo,,,kh u,kh ua,kh uah,kh uai,,kh uan,kh uann,kh uat,kh ue,kh ueh,kh uh,kh ui,kh uinn,kh un,kh ut
+g,g,g a,,g ai,,,g ak,g am,g an,g ang,,,,,g au,,,g e,,,,g i,g ia,g iah,,g iam,g ian,g iang,,,g iap,g iat,g iau,,,g ik,g im,g in,g ing,,g io,g ioh,g iok,g iong,,,,g iu,,,,,,,g o,,g ok,,g ong,,g oo,,,g u,g ua,,,,g uan,,g uat,g ue,g ueh,,g ui,,g uan,
+ng,ng,ng a,,ng ai,,,,,,,,,,,ng au,,,ng e,ng eh,,,,ng ia,,,,,,,,,,ng iau,ng iauh,,,,,,,,,,,,,,ng iu,,,,,,,,,,,,,ng oo,,,,,,,,,,,,,,,,,
+h,h,h a,h ah,h ai,h aih,h ainn,h ak,h am,h an,h ang,h ann,h annh,h ap,h at,h au,,,h e,h eh,,,h i,h ia,h iah,,h iam,h ian,h iang,h iann,h iannh,h iap,h iat,h iau,h iauh,,h ik,h im,h in,h ing,h inn,h io,h ioh,h iok,h iong,,h ip,h it,h iu,,h iunn,h m,h mh,h ng,h ngh,h o,h oh,h ok,,h ong,h onn,h oo,h ooh,,h u,h ua,h uah,h uai,h uainn,h uan,h uann,h uat,h ue,h ueh,,h ui,,h un,h ut
+ts,ts,ts a,ts ah,ts ai,,ts ainn,ts ak,ts am,ts an,ts ang,ts ann,,ts ap,ts at,ts au,,,ts e,ts eh,ts enn,,ts i,ts ia,ts iah,,ts iam,ts ian,ts iang,ts iann,,ts iap,ts iat,ts iau,,ts ih,ts ik,ts im,ts in,ts ing,ts inn,ts io,ts ioh,ts iok,ts iong,,ts ip,ts it,ts iu,ts iuh,ts iunn,,,ts ng,,ts o,ts oh,ts ok,,ts ong,,ts oo,,,ts u,ts ua,ts uah,,ts uainn,ts uan,ts uann,ts uat,ts ue,,ts uh,ts ui,,ts un,ts ut
+tsh,tsh,tsh a,tsh ah,tsh ai,,,tsh ak,tsh am,tsh an,tsh ang,tsh ann,,tsh ap,tsh at,tsh au,tsh auh,,tsh e,tsh eh,tsh enn,,tsh i,tsh ia,tsh iah,tsh iak,tsh iam,tsh ian,tsh iang,tsh iann,,tsh iap,tsh iat,tsh iau,,tsh ih,tsh ik,tsh im,tsh in,tsh ing,tsh inn,tsh io,tsh ioh,tsh iok,tsh iong,,tsh ip,tsh it,tsh iu,,tsh iunn,,,tsh ng,,tsh o,tsh oh,tsh ok,,tsh ong,,tsh oo,,,tsh u,tsh ua,tsh uah,,,tsh uan,tsh uann,,tsh ue,,tsh uh,tsh ui,,tsh un,tsh ut
+s,s,s a,s ah,s ai,,,s ak,s am,s an,s ang,s ann,s annh,s ap,s at,s au,,,s e,s eh,s enn,,s i,s ia,s iah,s iak,s iam,s ian,s iang,s iann,,s iap,s iat,s iau,,s ih,s ik,s im,s in,s ing,s inn,s io,s ioh,s iok,s iong,s iunn,s ip,s it,s iu,,s iunn,,,s ng,,s o,s oh,s ok,s om,s ong,,s oo,,,s u,s ua,s uah,,s uainn,s uan,s uann,s uat,s ue,s ueh,s uh,s ui,,s un,s ut
+j,j,,,,,,,,,,,,,,,,,,,,,j i,j ia,j iah, ,j iam,j ian,j iang,,,j iap,j iat,j iau,,,,j im,j in,,,j io,,j iok,j iong,,j ip,j it,j iu,,,,,,,,,,,,,,,,j u,,j uah,,,,,,j ue,,,,,j un,
\ No newline at end of file

utils

+../../../tools/kaldi/egs/wsj/s5/utils
\ No newline at end of file