testadmin
/
ParakeetEricRoss
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

add sclite test for text_frontend

This commit is contained in:
TianYuan 2021-08-21 08:35:25 +00:00
parent 5232f59840
commit 106891f443
3 changed files with 31 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
examples/text_frontend/run.sh
View File

@ -1,14 +1,37 @@
#!/bin/bash #!/bin/bash
USE_SCLITE=true
if [ "$USE_SCLITE" = true ];then
if [ ! -d "./SCTK" ];then
echo "Clone SCTK ..."
git clone https://github.com/usnistgov/SCTK
echo "Clone SCTK done!"
fi
if [ ! -d "./SCTK/bin" ];then
echo "Start make SCTK ..."
pushd SCTK && make config && make all && make check && make install && make doc && popd
echo "SCTK make done!"
fi
fi
# test g2p # test g2p
echo "Start get g2p test data." echo "Start get g2p test data ..."
python3 get_g2p_data.py --root-dir=~/datasets/BZNSYP --output-dir=data/g2p python3 get_g2p_data.py --root-dir=~/datasets/BZNSYP --output-dir=data/g2p
echo "Start test g2p." echo "Start test g2p ..."
python3 test_g2p.py --input-dir=data/g2p --output-dir=exp/g2p python3 test_g2p.py --input-dir=data/g2p --output-dir=exp/g2p
# test text normalization # test text normalization
echo "Start get text normalization test data." echo "Start get text normalization test data ..."
python3 get_textnorm_data.py --test-file=data/textnorm_test_cases.txt --output-dir=data/textnorm python3 get_textnorm_data.py --test-file=data/textnorm_test_cases.txt --output-dir=data/textnorm
echo "Start test text normalization." echo "Start test text normalization ..."
python3 test_textnorm.py --input-dir=data/textnorm --output-dir=exp/textnorm python3 test_textnorm.py --input-dir=data/textnorm --output-dir=exp/textnorm
# whether use sclite to get more detail information of WER
if [ "$USE_SCLITE" = true ];then
echo "Start sclite g2p ..."
./SCTK/bin/sclite -i wsj -r ./exp/g2p/text.ref.clean -h ./exp/g2p/text.g2p -e utf-8 -o all
echo
echo "Start sclite textnorm ..."
./SCTK/bin/sclite -i wsj -r ./exp/textnorm/text.ref.clean -h ./exp/textnorm/text.tn -e utf-8 -o all
fi

4
examples/text_frontend/test_g2p.py
View File

@ -47,8 +47,8 @@ def get_avg_wer(raw_dict, ref_dict, frontend, output_dir):
gt_phones = [phn for phn in gt_phones if phn not in SILENCE_TOKENS] gt_phones = [phn for phn in gt_phones if phn not in SILENCE_TOKENS]
gt_phones = " ".join(gt_phones) gt_phones = " ".join(gt_phones)
g2p_phones = " ".join(g2p_phones) g2p_phones = " ".join(g2p_phones)
wf_ref.write(utt_id + " " + gt_phones + "\n") wf_ref.write(gt_phones + "(baker_" + utt_id + ")" + "\n")
wf_g2p.write(utt_id + " " + g2p_phones + "\n") wf_g2p.write(g2p_phones + "(baker_" + utt_id + ")" + "\n")
edit_distance, ref_len = word_errors(gt_phones, g2p_phones) edit_distance, ref_len = word_errors(gt_phones, g2p_phones)
edit_distances.append(edit_distance) edit_distances.append(edit_distance)
ref_lens.append(ref_len) ref_lens.append(ref_len)

4
examples/text_frontend/test_textnorm.py
View File

@ -43,8 +43,8 @@ def get_avg_cer(raw_dict, ref_dict, text_normalizer, output_dir):
gt_text = del_en_add_space(gt_text) gt_text = del_en_add_space(gt_text)
textnorm_text = del_en_add_space(textnorm_text) textnorm_text = del_en_add_space(textnorm_text)
wf_ref.write(text_id + " " + gt_text + "\n") wf_ref.write(gt_text + "(" + text_id + ")" + "\n")
wf_tn.write(text_id + " " + textnorm_text + "\n") wf_tn.write(textnorm_text + "(" + text_id + ")" + "\n")
edit_distance, ref_len = char_errors(gt_text, textnorm_text) edit_distance, ref_len = char_errors(gt_text, textnorm_text)
edit_distances.append(edit_distance) edit_distances.append(edit_distance)
ref_lens.append(ref_len) ref_lens.append(ref_len)