wer.sh

#!/bin/bash

. ./path.sh || exit 1;

testset=$1
model=$2
chunk=$3

echo "*********"
echo "testset: ${testset}"
echo "model: ${model}"
echo "chunk: ${chunk}"
echo "*********"

test_dir=exp/${model}/test_${testset}
bpemodel=exp/wenet_efficient_conformer_librispeech_v2/train_960_unigram5000

for x in text_avg20_attention_${chunk} text_avg20_attention_rescoring_${chunk} text_avg20_ctc_prefix_beam_search_${chunk} text_avg20_ctc_greedy_search_${chunk} ; do
    echo "******************"
    echo "text file: ${x}"

    cut -f2- -d " " $test_dir/${x} > $test_dir/${x}_bpe_value_tmp
    cut -f1 -d " " $test_dir/${x} > $test_dir/${x}_bpe_key_tmp

    tools/spm_decode --model=${bpemodel}.model --input_format=piece \
      < $test_dir/${x}_bpe_value_tmp | sed -e "s/▁/ /g" > $test_dir/${x}_value_tmp

    paste -d " " $test_dir/${x}_bpe_key_tmp $test_dir/${x}_value_tmp > $test_dir/${x}

    python tools/compute-wer.py --char=1 --v=1 \
       data/${testset}/text ${test_dir}/$x > ${test_dir}/cer_$x.txt

    echo ${test_dir}/cer_$x.txt

done