diff --git a/docker/ajn_asr/main.py b/docker/ajn_asr/main.py
index 150ddde32759f53482c3f395601fe2f967d90b4c..bf69371d360bdde4dec3aa9b08af03b4b357a04b 100644
--- a/docker/ajn_asr/main.py
+++ b/docker/ajn_asr/main.py
@@ -7,12 +7,20 @@ from sziszapangma.integration.service_core.asr.asr_result import AsrResult
class SpeechbrainAsrProcessor(AsrBaseProcessor):
+ @staticmethod
+ def _process_file_to_correct_format(file_path: str, extension: str):
+ temp_file = f'{str(uuid.uuid4())}.{extension}'
+ os.system(f'ffmpeg -i {file_path} -ar 16000 {temp_file}')
+ # os.remove(file_path)
+ os.rename(temp_file, file_path)
+
def process_asr(self, audio_file_path: str) -> AsrResult:
file_tag = str(uuid.uuid4())
file_extension = audio_file_path.split('.')[-1]
file_name = f'{file_tag}.{file_extension}'
result_file_path = f'processing_flask/{file_tag}.txt'
file_path = f'processing_flask/{file_name}'
+ self._process_file_to_correct_format(audio_file_path, file_extension)
# create file in /data/uuid.ext
os.system(f"cp {audio_file_path} /data/{file_path}")
diff --git a/docker/ajn_asr/prepare_docker.sh b/docker/ajn_asr/prepare_docker.sh
index c7e687d087cba4d4c5fd7305b9ec8b6afb8fe431..711edd2c16d03bde15460d7f307599cb3d5bc109 100755
--- a/docker/ajn_asr/prepare_docker.sh
+++ b/docker/ajn_asr/prepare_docker.sh
@@ -3,5 +3,5 @@
SCRIPT_DIR="$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" &>/dev/null && pwd)"
docker build -t asr-clarin-pl-service "$SCRIPT_DIR"
-docker tag asr-clarin-pl-service gitlab.clarin-pl.eu:5050/clarin-dialog/clarin-dialog/asr-clarin-pl-service:1.4
-docker push gitlab.clarin-pl.eu:5050/clarin-dialog/clarin-dialog/asr-clarin-pl-service:1.4
+docker tag asr-clarin-pl-service gitlab.clarin-pl.eu:5050/clarin-dialog/clarin-dialog/asr-clarin-pl-service:1.5
+docker push gitlab.clarin-pl.eu:5050/clarin-dialog/clarin-dialog/asr-clarin-pl-service:1.5
diff --git a/docker/docker-compose.yml b/docker/docker-compose.yml
index d4486f1d65ce5621a00b1c80b6e63ae16a04e00e..b2fbf23ff98662ccaa5da524f131f7b9ebebd57a 100644
--- a/docker/docker-compose.yml
+++ b/docker/docker-compose.yml
@@ -18,34 +18,34 @@ services:
# - TECHMO_SERVER_URL=156.17.135.34
# - AUTH_TOKEN=__example_token__
- transformers-wav2vec2for_ctc:
- image: docker-registry.theliver.pl/transformers-wav2vec2for_ctc:1.0
- container_name: transformers-wav2vec2for_ctc
+ # transformers-wav2vec2for_ctc:
+ # image: gitlab.clarin-pl.eu:5050/clarin-dialog/clarin-dialog/transformers-wav2vec2for_ctc:1.0
+ # container_name: transformers-wav2vec2for_ctc
+ # restart: always
+ # volumes:
+ # - /etc/localtime:/etc/localtime:ro
+ # - ./wav2vec2for_ctc_models:/models
+ # ports:
+ # - "5430:5000"
+ # environment:
+ # - AUTH_TOKEN=__example_token__
+ # - MODEL_NAME=jonatasgrosman/wav2vec2-large-xlsr-53-polish
+ # - SAMPLING_RATE=16000
+
+ embedding_service:
+ image: docker-registry.theliver.pl/embedding_docker:1.0
+ container_name: embeddings_service
restart: always
- volumes:
- - /etc/localtime:/etc/localtime:ro
- - ./wav2vec2for_ctc_models:/models
ports:
- - "5430:5000"
+ - "5003:5000"
environment:
- - AUTH_TOKEN=__example_token__
- - MODEL_NAME=jonatasgrosman/wav2vec2-large-xlsr-53-polish
- - SAMPLING_RATE=16000
-
- # embedding_service:
- # image: docker-registry.theliver.pl/embedding_docker:1.0
- # container_name: embeddings_service
- # restart: always
- # ports:
- # - 5003:5000
- # environment:
- # - AUTH_TOKEN=__example_token__
- # volumes:
- # - /etc/localtime:/etc/localtime:ro
- # - ./embedding_models:/models
+ - AUTH_TOKEN=fjsd-mkwe-oius-m9h2
+ volumes:
+ - /etc/localtime:/etc/localtime:ro
+ - ./embedding_models:/models
ajn_asr:
- image: docker-registry.theliver.pl/asr-clarin-pl-service:1.4
+ image: gitlab.clarin-pl.eu:5050/clarin-dialog/clarin-dialog/asr-clarin-pl-service:1.5
container_name: ajn_asr
restart: always
ports:
@@ -55,14 +55,14 @@ services:
volumes:
- /etc/localtime:/etc/localtime:ro
- speechbrain_asr:
- image: docker-registry.theliver.pl/speechbrain-asr:1.5
- container_name: speechbrain_asr
- restart: always
- ports:
- - "5432:5000"
- volumes:
- - /etc/localtime:/etc/localtime:ro
- - ./speechbrain_asr_models:/models
- environment:
- - AUTH_TOKEN=__example_token__
+# speechbrain_asr:
+# image: gitlab.clarin-pl.eu:5050/clarin-dialog/clarin-dialog/speechbrain-asr:1.5
+# container_name: speechbrain_asr
+# restart: always
+# ports:
+# - "5432:5000"
+# volumes:
+# - /etc/localtime:/etc/localtime:ro
+# - ./speechbrain_asr_models:/models
+# environment:
+# - AUTH_TOKEN=__example_token__
diff --git a/dvc.lock b/dvc.lock
index 7a9261df1e54c57a45cf1d7152c4df94780b4dfd..85522fea163f790c37506b278f906fb85d4a4c8c 100644
--- a/dvc.lock
+++ b/dvc.lock
@@ -52,7 +52,7 @@ stages:
size: 229007155
nfiles: 1000
voicelab_import_to_common_format:
- cmd: PYTHONPATH=. python experiment/voicelab/import_data.py
+ cmd: PYTHONPATH=. python -u experiment/voicelab/import_data.py
deps:
- path: experiment/voicelab/import_data.py
md5: 41acb98a1517e66c052182fe0a1403ba
@@ -63,7 +63,7 @@ stages:
nfiles: 1600
outs:
- path: experiment_data/dataset_relation_manager_data/voicelab_cbiz_testset_20220322
- md5: 926ef9bab4ce41b9de95f2f3d5ab67a0.dir
+ md5: 4046ea5d80966f0c017b2c4bec0e7c9b.dir
size: 110711470
nfiles: 1600
luna_gold_transcript_processing:
@@ -93,8 +93,12 @@ stages:
cmd: "PYTHONPATH=. python experiment/luna/pipeline/luna_ajn_asr_processing.py\n"
deps:
- path: experiment/luna/pipeline/luna_ajn_asr_processing.py
- md5: ec7d7b5384f845173d9fb77e9cfa9907
- size: 2501
+ md5: 2d66cb8890c420b55e8b7eb33ac32ba2
+ size: 3558
+ - path: experiment_data/cached_asr/luna_ajn_polish_asr
+ md5: 620e178854dbcb69f49a608f34573a88.dir
+ size: 6159899
+ nfiles: 494
- path: experiment_data/dataset/LUNA.PL
md5: d342155b1871e881797cf7da09d5dc3c.dir
size: 1578358645
@@ -109,35 +113,43 @@ stages:
nfiles: 500
outs:
- path: experiment_data/pipeline/asr_benchmark_luna/ajn_polish_asr
- md5: 620e178854dbcb69f49a608f34573a88.dir
- size: 6159899
- nfiles: 494
+ md5: fa9d926ae8fd0268c71f19c1d5d39fcf.dir
+ size: 11080541
+ nfiles: 499
- path: experiment_data/pipeline/asr_benchmark_luna/ajn_spacy
- md5: 312be284d4ec9e38986048e785fcbbc1.dir
- size: 6535212
- nfiles: 494
+ md5: 417d8f07266eb5da9c4bfbf84f3b4eac.dir
+ size: 6579351
+ nfiles: 499
- path: experiment_data/pipeline/asr_benchmark_luna/pos_ajn_alignment_wer
- md5: 8ad558edb6a8bd2508a7e25bcf53bf94.dir
- size: 21936929
- nfiles: 494
+ md5: 2bf746c412e6bff4071f689d853b106f.dir
+ size: 22061350
+ nfiles: 499
- path: experiment_data/pipeline/asr_benchmark_luna/pos_ajn_metrics_wer
- md5: 98c74c5bf87637749eac1ed5ff3393b4.dir
- size: 16842
- nfiles: 494
+ md5: 3147413bdfd36ad91c64303e8705951b.dir
+ size: 17002
+ nfiles: 499
- path: experiment_data/pipeline/asr_benchmark_luna/word_ajn_alignment_wer
- md5: 1741fff740259398b28bf2a6ba3aec41.dir
- size: 20671277
- nfiles: 494
+ md5: 2bb11f8a97cdeb18c557fadb49a6f015.dir
+ size: 25669158
+ nfiles: 499
+ - path: experiment_data/pipeline/asr_benchmark_luna/word_ajn_alignment_wer_embeddings
+ md5: c2824c0c5cf433dbf864ebbdc2fb3cfc.dir
+ size: 44326962
+ nfiles: 500
- path: experiment_data/pipeline/asr_benchmark_luna/word_ajn_metrics_wer
- md5: 18605657ff9c7ef3221e27b671a3b4d1.dir
- size: 16835
- nfiles: 494
+ md5: c48c74eccf1cfd0768900514d2fcfd1b.dir
+ size: 10527
+ nfiles: 499
+ - path: experiment_data/pipeline/asr_benchmark_luna/word_ajn_metrics_wer_embeddings
+ md5: 98a7edeee3b630e8e301acfc578a8393.dir
+ size: 34869
+ nfiles: 500
luna_techmo_processing:
cmd: "PYTHONPATH=. python experiment/luna/pipeline/luna_techmo_processing.py\n"
deps:
- path: experiment/luna/pipeline/luna_techmo_processing.py
- md5: b4d5ad7a0d7fb0714a2dc02cb457e8c9
- size: 2628
+ md5: 75069cd6e3a61dfaaf49c2bdb1e81976
+ size: 3416
- path: experiment_data/cached_asr/luna_techmo
md5: 033ea7b5434dded73bf869bfdd299462.dir
size: 4256479
@@ -156,7 +168,7 @@ stages:
nfiles: 500
outs:
- path: experiment_data/pipeline/asr_benchmark_luna/pos_techmo_alignment_wer
- md5: c71539f3889c627a371957958bd0907d.dir
+ md5: 94762d19a853810064afd38319d05a2c.dir
size: 20897599
nfiles: 500
- path: experiment_data/pipeline/asr_benchmark_luna/pos_techmo_metrics_wer
@@ -164,21 +176,29 @@ stages:
size: 17341
nfiles: 500
- path: experiment_data/pipeline/asr_benchmark_luna/techmo_polish_asr
- md5: acfaec46b2415ed6a64e3a3464d164f8.dir
+ md5: 3787c6a4c7941787253165e2ba760e73.dir
size: 9697519
nfiles: 500
- path: experiment_data/pipeline/asr_benchmark_luna/techmo_spacy
- md5: e869581816457d1585a7e42d0a18b8b2.dir
+ md5: 337b6bf947ee47cda30b3cc75f954e8e.dir
size: 6124559
nfiles: 500
- path: experiment_data/pipeline/asr_benchmark_luna/word_techmo_alignment_wer
- md5: 0dabd65b3981d588cd23d943abc6e231.dir
+ md5: afc25d6ad22bed4ded5cb07028bff1cf.dir
size: 21380796
nfiles: 500
+ - path: experiment_data/pipeline/asr_benchmark_luna/word_techmo_alignment_wer_embeddings
+ md5: c2824c0c5cf433dbf864ebbdc2fb3cfc.dir
+ size: 44326962
+ nfiles: 500
- path: experiment_data/pipeline/asr_benchmark_luna/word_techmo_metrics_wer
md5: 4cfbb2830b280084ece14b1ef815b92a.dir
size: 17298
nfiles: 500
+ - path: experiment_data/pipeline/asr_benchmark_luna/word_techmo_metrics_wer_embeddings
+ md5: 98a7edeee3b630e8e301acfc578a8393.dir
+ size: 34869
+ nfiles: 500
voicelab_gold_transcript_processing:
cmd: "PYTHONPATH=. python -u experiment/voicelab/voicelab_pipeline_gold_transcript.py\n"
deps:
@@ -190,24 +210,24 @@ stages:
size: 4803739404
nfiles: 1600
- path: experiment_data/dataset_relation_manager_data/voicelab_cbiz_testset_20220322
- md5: 926ef9bab4ce41b9de95f2f3d5ab67a0.dir
+ md5: 4046ea5d80966f0c017b2c4bec0e7c9b.dir
size: 110711470
nfiles: 1600
outs:
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript
- md5: fb6812b2f3044c0285ee6ee2b21d0523.dir
+ md5: 9edf1e743faa9fc3515790acb6fd8cab.dir
size: 21846798
nfiles: 800
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript_spacy
- md5: f2e68dcc8842a15e417ae6f5221a802a.dir
+ md5: c166937f6e8ae9d28412ca1e3e43469e.dir
size: 26643278
nfiles: 800
voicelab_techmo_processing:
cmd: "PYTHONPATH=. python -u experiment/voicelab/voicelab_pipeline_techmo.py\n"
deps:
- path: experiment/voicelab/voicelab_pipeline_techmo.py
- md5: 23c0869d7cc9f0088870362d669ab82e
- size: 2685
+ md5: 3d6347486055a11e399beac71ce2f877
+ size: 3479
- path: experiment_data/cached_asr/voicelab_cbiz_testset_20220322_techmo
md5: 6c3b356723d562c978f84e733b91f5d0.dir
size: 17539259
@@ -217,16 +237,16 @@ stages:
size: 4803739404
nfiles: 1600
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript
- md5: fb6812b2f3044c0285ee6ee2b21d0523.dir
+ md5: 9edf1e743faa9fc3515790acb6fd8cab.dir
size: 21846798
nfiles: 800
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript_spacy
- md5: f2e68dcc8842a15e417ae6f5221a802a.dir
+ md5: c166937f6e8ae9d28412ca1e3e43469e.dir
size: 26643278
nfiles: 800
outs:
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_techmo_alignment_wer
- md5: 8c5f0380ba2891b3e726d647c2863c60.dir
+ md5: a15a7a19f46e329c8b77eeecdda9d7b4.dir
size: 81650836
nfiles: 800
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_techmo_metrics_wer
@@ -234,18 +254,81 @@ stages:
size: 27934
nfiles: 800
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/techmo_polish_asr
- md5: c45e29b08af7bb13cdf54da9655bd96c.dir
+ md5: da32e6fa9d986deddb594cb66e649864.dir
size: 39158267
nfiles: 800
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/techmo_spacy
- md5: a39c82666419c2b7791952a1fa116d61.dir
+ md5: cd89a91a33629088ba6fc30ef8427dee.dir
size: 24482297
nfiles: 800
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_techmo_alignment_wer
- md5: 72ff86c7cb2e89ac7e04677f532255b2.dir
+ md5: 0b714391682432408d74beee1cd5a14a.dir
size: 83756423
nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_techmo_alignment_wer_embeddings
+ md5: 93d34d82f8536014ddbe0cf0645dd837.dir
+ size: 174322727
+ nfiles: 800
- path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_techmo_metrics_wer
md5: 2fe3288abe85e4a385e2aefa0e8cad7e.dir
size: 27780
nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_techmo_metrics_wer_embeddings
+ md5: 1fc2985ad4c3cb00d05b1865ad5b22d4.dir
+ size: 56182
+ nfiles: 800
+ voicelab_ajn_processing:
+ cmd: "PYTHONPATH=. python -u experiment/voicelab/voicelab_pipeline_ajn_asr.py\n"
+ deps:
+ - path: experiment/voicelab/voicelab_pipeline_ajn_asr.py
+ md5: 85e8d3d79379e6d5db751e03c5ebae75
+ size: 4161
+ - path: experiment_data/cached_asr/voicelab_cbiz_testset_20220322_ajn
+ md5: 49a38b90f1265a61b90b54f820415011.dir
+ size: 32601414
+ nfiles: 800
+ - path: experiment_data/dataset/voicelab_cbiz_testset_20220322
+ md5: 3c2b18e1f1f89e4c5ad7b254e472b25e.dir
+ size: 4803739404
+ nfiles: 1600
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript
+ md5: 9edf1e743faa9fc3515790acb6fd8cab.dir
+ size: 21846798
+ nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript_spacy
+ md5: c166937f6e8ae9d28412ca1e3e43469e.dir
+ size: 26643278
+ nfiles: 800
+ outs:
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/ajn_polish_asr
+ md5: 94181d7a0731e8defbdcb4b477ad72a2.dir
+ size: 48470646
+ nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/ajn_spacy
+ md5: ef8be18b8acca299f9b9542ac8643a87.dir
+ size: 20536889
+ nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_ajn_alignment_wer
+ md5: b2d3a9872e6016cfde8e6d025bef373b.dir
+ size: 78539613
+ nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_ajn_metrics_wer
+ md5: d0e1ef5f57de27a2356d2f2050a93349.dir
+ size: 27353
+ nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_alignment_wer
+ md5: acb5337346e70bed974dfe7ca7947d79.dir
+ size: 104789466
+ nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_alignment_wer_embeddings
+ md5: 93d34d82f8536014ddbe0cf0645dd837.dir
+ size: 174322727
+ nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_metrics_wer
+ md5: 903096554a3ea6896c4abaa5e2c71d4c.dir
+ size: 16505
+ nfiles: 800
+ - path: experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_metrics_wer_embeddings
+ md5: 1fc2985ad4c3cb00d05b1865ad5b22d4.dir
+ size: 56182
+ nfiles: 800
diff --git a/dvc.yaml b/dvc.yaml
index da6552da4616572e8a7680ea957eb585a1a11fe6..e2f3e5d29374acdce61d0676cd4afe8f77add483 100644
--- a/dvc.yaml
+++ b/dvc.yaml
@@ -35,6 +35,8 @@ stages:
- experiment_data/pipeline/asr_benchmark_luna/techmo_spacy
- experiment_data/pipeline/asr_benchmark_luna/pos_techmo_alignment_wer
- experiment_data/pipeline/asr_benchmark_luna/pos_techmo_metrics_wer
+ - experiment_data/pipeline/asr_benchmark_luna/word_techmo_metrics_wer_embeddings
+ - experiment_data/pipeline/asr_benchmark_luna/word_techmo_alignment_wer_embeddings
luna_ajn_processing:
cmd: |
@@ -42,6 +44,7 @@ stages:
deps:
- experiment/luna/pipeline/luna_ajn_asr_processing.py
- experiment_data/dataset/LUNA.PL
+ - experiment_data/cached_asr/luna_ajn_polish_asr
- experiment_data/pipeline/asr_benchmark_luna/gold_transcript
- experiment_data/pipeline/asr_benchmark_luna/gold_transcript_spacy
outs:
@@ -51,9 +54,11 @@ stages:
- experiment_data/pipeline/asr_benchmark_luna/ajn_spacy
- experiment_data/pipeline/asr_benchmark_luna/pos_ajn_alignment_wer
- experiment_data/pipeline/asr_benchmark_luna/pos_ajn_metrics_wer
+ - experiment_data/pipeline/asr_benchmark_luna/word_ajn_metrics_wer_embeddings
+ - experiment_data/pipeline/asr_benchmark_luna/word_ajn_alignment_wer_embeddings
voicelab_import_to_common_format:
- cmd: PYTHONPATH=. python experiment/voicelab/import_data.py
+ cmd: PYTHONPATH=. python -u experiment/voicelab/import_data.py
deps:
- experiment/voicelab/import_data.py
- experiment_data/dataset/voicelab_cbiz_testset_20220322
@@ -87,21 +92,26 @@ stages:
- experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/techmo_spacy
- experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_techmo_alignment_wer
- experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_techmo_metrics_wer
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_techmo_metrics_wer_embeddings
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_techmo_alignment_wer_embeddings
-# voicelab_ajn_processing:
-# cmd: |
-# PYTHONPATH=. python -u experiment/voicelab/voicelab_pipeline_ajn_asr.py
-# deps:
-# - experiment/voicelab/voicelab_pipeline_ajn_asr.py
-# - experiment_data/dataset/voicelab_cbiz_testset_20220322
-# - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript
-# - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript_spacy
-# outs:
-# - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/ajn_polish_asr
-# - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_metrics_wer
-# - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_alignment_wer
-# - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/ajn_spacy
-# - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_ajn_alignment_wer
-# - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_ajn_metrics_wer
+ voicelab_ajn_processing:
+ cmd: |
+ PYTHONPATH=. python -u experiment/voicelab/voicelab_pipeline_ajn_asr.py
+ deps:
+ - experiment/voicelab/voicelab_pipeline_ajn_asr.py
+ - experiment_data/cached_asr/voicelab_cbiz_testset_20220322_ajn
+ - experiment_data/dataset/voicelab_cbiz_testset_20220322
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/gold_transcript_spacy
+ outs:
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/ajn_polish_asr
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_metrics_wer
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_alignment_wer
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/ajn_spacy
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_ajn_alignment_wer
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/pos_ajn_metrics_wer
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_metrics_wer_embeddings
+ - experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/word_ajn_alignment_wer_embeddings
# concurrent features, multiprocessing
diff --git a/experiment/luna/pipeline/dependency_provider.py b/experiment/luna/pipeline/dependency_provider.py
index 7557f511e1585a1d5b443ca181a607e9595e60c2..78a4e7ab05cf82d1de8980e97905958b0210bab9 100644
--- a/experiment/luna/pipeline/dependency_provider.py
+++ b/experiment/luna/pipeline/dependency_provider.py
@@ -8,18 +8,22 @@ GOLD_TRANSCRIPT = 'gold_transcript'
GOLD_TRANSCRIPT_SPACY = 'gold_transcript_spacy'
TECHMO_POLISH_ASR = 'techmo_polish_asr'
-WORD_TECHMO_MERTICS_WER = 'word_techmo_metrics_wer'
+WORD_TECHMO_METRICS_WER = 'word_techmo_metrics_wer'
WORD_TECHMO_ALIGNMENT_WER = 'word_techmo_alignment_wer'
TECHMO_SPACY = 'techmo_spacy'
POS_TECHMO_ALIGNMENT_WER = 'pos_techmo_alignment_wer'
POS_TECHMO_METRICS_WER = 'pos_techmo_metrics_wer'
+WORD_TECHMO_METRICS_WER_EMBEDDINGS = 'word_techmo_metrics_wer_embeddings'
+WORD_TECHMO_ALIGNMENT_WER_EMBEDDINGS = 'word_techmo_alignment_wer_embeddings'
AJN_POLISH_ASR = 'ajn_polish_asr'
-WORD_AJN_MERTICS_WER = 'word_ajn_metrics_wer'
+WORD_AJN_METRICS_WER = 'word_ajn_metrics_wer'
WORD_AJN_ALIGNMENT_WER = 'word_ajn_alignment_wer'
AJN_SPACY = 'ajn_spacy'
POS_AJN_ALIGNMENT_WER = 'pos_ajn_alignment_wer'
POS_AJN_METRICS_WER = 'pos_ajn_metrics_wer'
+WORD_AJN_METRICS_WER_EMBEDDINGS = 'word_ajn_metrics_wer_embeddings'
+WORD_AJN_ALIGNMENT_WER_EMBEDDINGS = 'word_ajn_alignment_wer_embeddings'
def get_record_provider() -> LunaRecordProvider:
diff --git a/experiment/luna/pipeline/luna_ajn_asr_processing.py b/experiment/luna/pipeline/luna_ajn_asr_processing.py
index 7ad6122475444b57eb1f209bf31945bcf7277ad8..662858dd2b6e1bfdfb706717e754b9259ac6d21f 100644
--- a/experiment/luna/pipeline/luna_ajn_asr_processing.py
+++ b/experiment/luna/pipeline/luna_ajn_asr_processing.py
@@ -1,13 +1,15 @@
from experiment.luna.pipeline.dependency_provider import get_record_provider, GOLD_TRANSCRIPT, \
get_multiple_files_repository, \
- GOLD_TRANSCRIPT_SPACY, AJN_POLISH_ASR, WORD_AJN_MERTICS_WER, WORD_AJN_ALIGNMENT_WER, AJN_SPACY, \
- POS_AJN_ALIGNMENT_WER, POS_AJN_METRICS_WER
+ GOLD_TRANSCRIPT_SPACY, AJN_POLISH_ASR, WORD_AJN_METRICS_WER, WORD_AJN_ALIGNMENT_WER, AJN_SPACY, \
+ POS_AJN_ALIGNMENT_WER, POS_AJN_METRICS_WER, WORD_AJN_METRICS_WER_EMBEDDINGS, WORD_AJN_ALIGNMENT_WER_EMBEDDINGS
from experiment.pos_processing.asr_spacy_token_pos_processing_task import AsrSpacyTokenPosProcessingTask
from experiment.pos_processing.spacy_pos_wer_processing_task import SpacyPosWerProcessingTask
+from sziszapangma.core.transformer.web_embedding_transformer import WebEmbeddingTransformer
from sziszapangma.integration.experiment_manager import ExperimentManager
from sziszapangma.integration.task.asr_task import AsrTask
from sziszapangma.integration.task.classic_wer_metric_task import ClassicWerMetricTask
-from sziszapangma.integration.asr_processor import AsrWebClient
+from sziszapangma.integration.asr_processor import AsrWebClient, MultipleSourcesAsrProcessor, AsrPathCacheClient
+from sziszapangma.integration.task.embedding_wer_metrics_task import EmbeddingWerMetricsTask
def run_luna_experiment():
@@ -17,7 +19,11 @@ def run_luna_experiment():
processing_tasks=[
AsrTask(
task_name='ajn_polish_asr_task',
- asr_processor=AsrWebClient('http://localhost:5431/process_asr', '__example_token__'),
+ asr_processor=MultipleSourcesAsrProcessor([
+ AsrPathCacheClient('experiment_data/cached_asr/luna_ajn_polish_asr', record_provider,
+ record_provider),
+ AsrWebClient('http://localhost:5431/process_asr', '__example_token__')
+ ]),
asr_property_name=AJN_POLISH_ASR,
require_update=False,
record_path_provider=record_provider
@@ -26,7 +32,7 @@ def run_luna_experiment():
task_name='techmo_word_wer_processing',
asr_property_name=AJN_POLISH_ASR,
gold_transcript_property_name=GOLD_TRANSCRIPT,
- metrics_property_name=WORD_AJN_MERTICS_WER,
+ metrics_property_name=WORD_AJN_METRICS_WER,
require_update=False,
alignment_property_name=WORD_AJN_ALIGNMENT_WER
),
@@ -43,6 +49,15 @@ def run_luna_experiment():
asr_pos_property_name=AJN_SPACY,
pos_alignment_wer=POS_AJN_ALIGNMENT_WER,
pos_metrics_wer=POS_AJN_METRICS_WER
+ ),
+ EmbeddingWerMetricsTask(
+ task_name='EmbeddingWerMetricsTask',
+ asr_property_name='techmo_polish_asr',
+ gold_transcript_property_name=GOLD_TRANSCRIPT,
+ metrics_property_name=WORD_AJN_METRICS_WER_EMBEDDINGS,
+ require_update=False,
+ embedding_transformer=WebEmbeddingTransformer('pl', 'http://localhost:5003', 'fjsd-mkwe-oius-m9h2'),
+ alignment_property_name=WORD_AJN_ALIGNMENT_WER_EMBEDDINGS
)
],
experiment_repository=get_multiple_files_repository(),
diff --git a/experiment/luna/pipeline/luna_techmo_processing.py b/experiment/luna/pipeline/luna_techmo_processing.py
index ee1229687f1a20903267ee015a15b8dd6e42124a..c219440bbd0893443d4014e8f891ae5b9efa84ea 100644
--- a/experiment/luna/pipeline/luna_techmo_processing.py
+++ b/experiment/luna/pipeline/luna_techmo_processing.py
@@ -1,13 +1,15 @@
from experiment.luna.pipeline.dependency_provider import get_record_provider, GOLD_TRANSCRIPT, TECHMO_POLISH_ASR, \
get_multiple_files_repository, \
- GOLD_TRANSCRIPT_SPACY, POS_TECHMO_ALIGNMENT_WER, POS_TECHMO_METRICS_WER, WORD_TECHMO_MERTICS_WER, \
- WORD_TECHMO_ALIGNMENT_WER, TECHMO_SPACY
+ GOLD_TRANSCRIPT_SPACY, POS_TECHMO_ALIGNMENT_WER, POS_TECHMO_METRICS_WER, WORD_TECHMO_METRICS_WER, \
+ WORD_TECHMO_ALIGNMENT_WER, TECHMO_SPACY, WORD_TECHMO_METRICS_WER_EMBEDDINGS, WORD_TECHMO_ALIGNMENT_WER_EMBEDDINGS
from experiment.pos_processing.asr_spacy_token_pos_processing_task import AsrSpacyTokenPosProcessingTask
from experiment.pos_processing.spacy_pos_wer_processing_task import SpacyPosWerProcessingTask
+from sziszapangma.core.transformer.web_embedding_transformer import WebEmbeddingTransformer
from sziszapangma.integration.asr_processor import AsrPathCacheClient
from sziszapangma.integration.experiment_manager import ExperimentManager
from sziszapangma.integration.task.asr_task import AsrTask
from sziszapangma.integration.task.classic_wer_metric_task import ClassicWerMetricTask
+from sziszapangma.integration.task.embedding_wer_metrics_task import EmbeddingWerMetricsTask
def run_luna_experiment():
@@ -27,7 +29,7 @@ def run_luna_experiment():
task_name='techmo_word_wer_processing',
asr_property_name=TECHMO_POLISH_ASR,
gold_transcript_property_name=GOLD_TRANSCRIPT,
- metrics_property_name=WORD_TECHMO_MERTICS_WER,
+ metrics_property_name=WORD_TECHMO_METRICS_WER,
require_update=False,
alignment_property_name=WORD_TECHMO_ALIGNMENT_WER
),
@@ -44,6 +46,15 @@ def run_luna_experiment():
asr_pos_property_name=TECHMO_SPACY,
pos_alignment_wer=POS_TECHMO_ALIGNMENT_WER,
pos_metrics_wer=POS_TECHMO_METRICS_WER
+ ),
+ EmbeddingWerMetricsTask(
+ task_name='EmbeddingWerMetricsTask',
+ asr_property_name='techmo_polish_asr',
+ gold_transcript_property_name=GOLD_TRANSCRIPT,
+ metrics_property_name=WORD_TECHMO_METRICS_WER_EMBEDDINGS,
+ require_update=False,
+ embedding_transformer=WebEmbeddingTransformer('pl', 'http://localhost:5003', 'fjsd-mkwe-oius-m9h2'),
+ alignment_property_name=WORD_TECHMO_ALIGNMENT_WER_EMBEDDINGS
)
],
experiment_repository=get_multiple_files_repository(),
diff --git a/experiment/voicelab/voicelab_dependency.py b/experiment/voicelab/voicelab_dependency.py
index 58642e403a4c2f0bb7a7db59d5013723e01453d2..4b528c9c2689c8457186c72e0a9a497202bbe369 100644
--- a/experiment/voicelab/voicelab_dependency.py
+++ b/experiment/voicelab/voicelab_dependency.py
@@ -8,19 +8,22 @@ GOLD_TRANSCRIPT = 'gold_transcript'
GOLD_TRANSCRIPT_SPACY = 'gold_transcript_spacy'
TECHMO_POLISH_ASR = 'techmo_polish_asr'
-WORD_TECHMO_MERTICS_WER = 'word_techmo_metrics_wer'
+WORD_TECHMO_METRICS_WER = 'word_techmo_metrics_wer'
WORD_TECHMO_ALIGNMENT_WER = 'word_techmo_alignment_wer'
TECHMO_SPACY = 'techmo_spacy'
POS_TECHMO_ALIGNMENT_WER = 'pos_techmo_alignment_wer'
POS_TECHMO_METRICS_WER = 'pos_techmo_metrics_wer'
+WORD_TECHMO_METRICS_WER_EMBEDDINGS = 'word_techmo_metrics_wer_embeddings'
+WORD_TECHMO_ALIGNMENT_WER_EMBEDDINGS = 'word_techmo_alignment_wer_embeddings'
AJN_POLISH_ASR = 'ajn_polish_asr'
-WORD_AJN_MERTICS_WER = 'word_ajn_metrics_wer'
+WORD_AJN_METRICS_WER = 'word_ajn_metrics_wer'
WORD_AJN_ALIGNMENT_WER = 'word_ajn_alignment_wer'
AJN_SPACY = 'ajn_spacy'
POS_AJN_ALIGNMENT_WER = 'pos_ajn_alignment_wer'
POS_AJN_METRICS_WER = 'pos_ajn_metrics_wer'
-
+WORD_AJN_METRICS_WER_EMBEDDINGS = 'word_ajn_metrics_wer_embeddings'
+WORD_AJN_ALIGNMENT_WER_EMBEDDINGS = 'word_ajn_alignment_wer_embeddings'
PIPELINE_DATA_DIRECTORY = 'experiment_data/pipeline'
EXPERIMENT_NAME = 'asr_benchmark_voicelab_cbiz_testset_20220322'
diff --git a/experiment/voicelab/voicelab_pipeline.py b/experiment/voicelab/voicelab_pipeline.py
deleted file mode 100644
index d58f92e6e14816641c67dc6e03fc8a35373c549b..0000000000000000000000000000000000000000
--- a/experiment/voicelab/voicelab_pipeline.py
+++ /dev/null
@@ -1,96 +0,0 @@
-from experiment.pos_processing.asr_spacy_token_pos_processing_task import AsrSpacyTokenPosProcessingTask
-from experiment.pos_processing.gold_transcript_spacy_token_pos_processing_task import \
- GoldTranscriptSpacyTokenPosProcessingTask
-from experiment.pos_processing.spacy_pos_wer_processing_task import SpacyPosWerProcessingTask
-from experiment.voicelab.voicelab_dependency import get_record_provider, get_repository
-from experiment.voicelab.voicelab_gold_transcript_processor import VoicelabGoldTranscriptProcessor
-from experiment.voicelab.voicelab_telco_record_provider import VoicelabTelcoRecordProvider
-from sziszapangma.integration.asr_processor import AsrPathCacheClient
-from sziszapangma.integration.experiment_manager import ExperimentManager
-from sziszapangma.integration.path_filter import ExtensionPathFilter
-from sziszapangma.integration.repository.experiment_repository import ExperimentRepository
-from sziszapangma.integration.repository.multi_files_experiment_repository import \
- MultiFilesExperimentRepository
-from sziszapangma.integration.task.asr_task import AsrTask
-from sziszapangma.integration.task.classic_wer_metric_task import ClassicWerMetricTask
-from sziszapangma.integration.task.gold_transcript_task import GoldTranscriptTask
-
-GOLD_TRANSCRIPT = 'gold_transcript'
-TECHMO_POLISH_ASR = 'techmo_polish_asr'
-TECHMO_POLISH_CLASSIC_WER_METRIC = 'techmo_polish_classic_wer_metric'
-TECHMO_POLISH_CLASSIC_ALIGNMENT = 'techmo_polish_classic_alignment'
-TECHMO_SPACY = 'techmo_spacy'
-GOLD_TRANSCRIPT_SPACY = 'gold_transcript_spacy'
-POS_ALIGNMENT_WER = 'pos_alignment_wer'
-POS_METRICS_WER = 'pos_metrics_wer'
-
-DATASET_DIRECTORY = 'experiment_data/dataset/voicelab_cbiz_testset_20220322'
-PIPELINE_DATA_DIRECTORY = 'experiment_data/pipeline'
-EXPERIMENT_NAME = 'asr_benchmark_voicelab_cbiz_testset_20220322'
-RELATION_MANAGER_ROOT_PATH = 'experiment_data/dataset_relation_manager_data/voicelab_cbiz_testset_20220322'
-
-
-def run_voicelab_experiment():
- record_provider = get_record_provider()
- experiment_processor = ExperimentManager(
- record_id_iterator=record_provider,
- processing_tasks=[
- GoldTranscriptTask(
- task_name='gold_transcript_task',
- gold_transcript_processor=VoicelabGoldTranscriptProcessor(record_provider),
- gold_transcript_property_name=GOLD_TRANSCRIPT,
- require_update=False
- ),
- AsrTask(
- task_name='techmo_polish_task',
- # asr_processor=AsrWebClient('http://192.168.0.124:4999/process_asr', 'test1234'),
- asr_processor=AsrPathCacheClient(
- 'experiment_data/cached_asr/voicelab_cbiz_testset_20220322_techmo',
- record_provider,
- record_provider
- ),
- asr_property_name=TECHMO_POLISH_ASR,
- require_update=False,
- record_path_provider=record_provider
- ),
- ClassicWerMetricTask(
- task_name='classic_wer_metric_task',
- asr_property_name=TECHMO_POLISH_ASR,
- gold_transcript_property_name=GOLD_TRANSCRIPT,
- metrics_property_name=TECHMO_POLISH_CLASSIC_WER_METRIC,
- require_update=False,
- alignment_property_name=TECHMO_POLISH_CLASSIC_ALIGNMENT
- ),
- GoldTranscriptSpacyTokenPosProcessingTask(
- task_name='gold_transcript_spacy_task',
- input_property_name=GOLD_TRANSCRIPT,
- spacy_property_name=GOLD_TRANSCRIPT_SPACY,
- require_update=True
- ),
- AsrSpacyTokenPosProcessingTask(
- task_name='techmo_spacy_task',
- input_property_name=TECHMO_POLISH_ASR,
- spacy_property_name=TECHMO_SPACY,
- require_update=True
- ),
- SpacyPosWerProcessingTask(
- task_name='PosWerProcessor',
- require_update=False,
- gold_transcript_pos_property_name=GOLD_TRANSCRIPT_SPACY,
- asr_pos_property_name=TECHMO_SPACY,
- pos_alignment_wer=POS_ALIGNMENT_WER,
- pos_metrics_wer=POS_METRICS_WER
- )
- ],
- experiment_repository=get_repository(),
- relation_manager_provider=record_provider
- )
- experiment_processor.process()
-
-
-def example_run():
- run_voicelab_experiment()
-
-
-if __name__ == '__main__':
- example_run()
diff --git a/experiment/voicelab/voicelab_pipeline_ajn_asr.py b/experiment/voicelab/voicelab_pipeline_ajn_asr.py
index d650082aaa8210b16f35d5d0075a919f30241e64..b6c4c2c441b8e1681d2418312de745ef1588c12b 100644
--- a/experiment/voicelab/voicelab_pipeline_ajn_asr.py
+++ b/experiment/voicelab/voicelab_pipeline_ajn_asr.py
@@ -3,11 +3,12 @@ from experiment.pos_processing.gold_transcript_spacy_token_pos_processing_task i
GoldTranscriptSpacyTokenPosProcessingTask
from experiment.pos_processing.spacy_pos_wer_processing_task import SpacyPosWerProcessingTask
from experiment.voicelab.voicelab_dependency import get_record_provider, get_repository, GOLD_TRANSCRIPT, \
- GOLD_TRANSCRIPT_SPACY, TECHMO_POLISH_ASR, AJN_POLISH_ASR, WORD_AJN_MERTICS_WER, WORD_AJN_ALIGNMENT_WER, AJN_SPACY, \
- POS_AJN_ALIGNMENT_WER, POS_AJN_METRICS_WER
+ GOLD_TRANSCRIPT_SPACY, TECHMO_POLISH_ASR, AJN_POLISH_ASR, WORD_AJN_METRICS_WER, WORD_AJN_ALIGNMENT_WER, AJN_SPACY, \
+ POS_AJN_ALIGNMENT_WER, POS_AJN_METRICS_WER, WORD_AJN_METRICS_WER_EMBEDDINGS, WORD_AJN_ALIGNMENT_WER_EMBEDDINGS
from experiment.voicelab.voicelab_gold_transcript_processor import VoicelabGoldTranscriptProcessor
from experiment.voicelab.voicelab_telco_record_provider import VoicelabTelcoRecordProvider
-from sziszapangma.integration.asr_processor import AsrPathCacheClient, AsrWebClient
+from sziszapangma.core.transformer.web_embedding_transformer import WebEmbeddingTransformer
+from sziszapangma.integration.asr_processor import AsrPathCacheClient, AsrWebClient, MultipleSourcesAsrProcessor
from sziszapangma.integration.experiment_manager import ExperimentManager
from sziszapangma.integration.path_filter import ExtensionPathFilter
from sziszapangma.integration.repository.experiment_repository import ExperimentRepository
@@ -15,6 +16,7 @@ from sziszapangma.integration.repository.multi_files_experiment_repository impor
MultiFilesExperimentRepository
from sziszapangma.integration.task.asr_task import AsrTask
from sziszapangma.integration.task.classic_wer_metric_task import ClassicWerMetricTask
+from sziszapangma.integration.task.embedding_wer_metrics_task import EmbeddingWerMetricsTask
from sziszapangma.integration.task.gold_transcript_task import GoldTranscriptTask
@@ -25,7 +27,11 @@ def run_voicelab_experiment():
processing_tasks=[
AsrTask(
task_name='ajn_polish_asr_task',
- asr_processor=AsrWebClient('http://localhost:5431/process_asr', '__example_token__'),
+ asr_processor=AsrPathCacheClient(
+ 'experiment_data/cached_asr/voicelab_cbiz_testset_20220322_ajn',
+ record_provider,
+ record_provider
+ ),
asr_property_name=AJN_POLISH_ASR,
require_update=False,
record_path_provider=record_provider
@@ -34,7 +40,7 @@ def run_voicelab_experiment():
task_name='techmo_word_wer_processing',
asr_property_name=AJN_POLISH_ASR,
gold_transcript_property_name=GOLD_TRANSCRIPT,
- metrics_property_name=WORD_AJN_MERTICS_WER,
+ metrics_property_name=WORD_AJN_METRICS_WER,
require_update=False,
alignment_property_name=WORD_AJN_ALIGNMENT_WER
),
@@ -51,6 +57,15 @@ def run_voicelab_experiment():
asr_pos_property_name=AJN_SPACY,
pos_alignment_wer=POS_AJN_ALIGNMENT_WER,
pos_metrics_wer=POS_AJN_METRICS_WER
+ ),
+ EmbeddingWerMetricsTask(
+ task_name='EmbeddingWerMetricsTask',
+ asr_property_name='techmo_polish_asr',
+ gold_transcript_property_name=GOLD_TRANSCRIPT,
+ metrics_property_name=WORD_AJN_METRICS_WER_EMBEDDINGS,
+ require_update=False,
+ embedding_transformer=WebEmbeddingTransformer('pl', 'http://localhost:5003', 'fjsd-mkwe-oius-m9h2'),
+ alignment_property_name=WORD_AJN_ALIGNMENT_WER_EMBEDDINGS
)
],
experiment_repository=get_repository(),
@@ -64,6 +79,4 @@ def example_run():
if __name__ == '__main__':
- # example_run()
- path = '/home/marcinwatroba/PWR_ASR/asr-benchmarks/experiment_data/dataset/voicelab_cbiz_testset_20220322/bankowe/cbiz_tc_2.agnt.wav'
- print(AsrWebClient('http://localhost:5431/process_asr', '__example_token__').call_recognise(path))
+ example_run()
diff --git a/experiment/voicelab/voicelab_pipeline_techmo.py b/experiment/voicelab/voicelab_pipeline_techmo.py
index 03bdd64f82f71b0230c93d095f73dc6774eae0b1..c99e09b1b95e5a33d19d7857289c6c966eabacd7 100644
--- a/experiment/voicelab/voicelab_pipeline_techmo.py
+++ b/experiment/voicelab/voicelab_pipeline_techmo.py
@@ -1,12 +1,15 @@
from experiment.pos_processing.asr_spacy_token_pos_processing_task import AsrSpacyTokenPosProcessingTask
from experiment.pos_processing.spacy_pos_wer_processing_task import SpacyPosWerProcessingTask
from experiment.voicelab.voicelab_dependency import get_record_provider, get_repository, GOLD_TRANSCRIPT, \
- GOLD_TRANSCRIPT_SPACY, TECHMO_POLISH_ASR, WORD_TECHMO_MERTICS_WER, WORD_TECHMO_ALIGNMENT_WER, TECHMO_SPACY, \
- POS_TECHMO_METRICS_WER, POS_TECHMO_ALIGNMENT_WER
+ GOLD_TRANSCRIPT_SPACY, TECHMO_POLISH_ASR, WORD_TECHMO_METRICS_WER, WORD_TECHMO_ALIGNMENT_WER, TECHMO_SPACY, \
+ POS_TECHMO_METRICS_WER, POS_TECHMO_ALIGNMENT_WER, WORD_TECHMO_METRICS_WER_EMBEDDINGS, \
+ WORD_TECHMO_ALIGNMENT_WER_EMBEDDINGS
+from sziszapangma.core.transformer.web_embedding_transformer import WebEmbeddingTransformer
from sziszapangma.integration.asr_processor import AsrPathCacheClient
from sziszapangma.integration.experiment_manager import ExperimentManager
from sziszapangma.integration.task.asr_task import AsrTask
from sziszapangma.integration.task.classic_wer_metric_task import ClassicWerMetricTask
+from sziszapangma.integration.task.embedding_wer_metrics_task import EmbeddingWerMetricsTask
def run_voicelab_experiment():
@@ -29,7 +32,7 @@ def run_voicelab_experiment():
task_name='techmo_word_wer_processing',
asr_property_name=TECHMO_POLISH_ASR,
gold_transcript_property_name=GOLD_TRANSCRIPT,
- metrics_property_name=WORD_TECHMO_MERTICS_WER,
+ metrics_property_name=WORD_TECHMO_METRICS_WER,
require_update=False,
alignment_property_name=WORD_TECHMO_ALIGNMENT_WER
),
@@ -46,6 +49,15 @@ def run_voicelab_experiment():
asr_pos_property_name=TECHMO_SPACY,
pos_alignment_wer=POS_TECHMO_ALIGNMENT_WER,
pos_metrics_wer=POS_TECHMO_METRICS_WER
+ ),
+ EmbeddingWerMetricsTask(
+ task_name='EmbeddingWerMetricsTask',
+ asr_property_name='techmo_polish_asr',
+ gold_transcript_property_name=GOLD_TRANSCRIPT,
+ metrics_property_name=WORD_TECHMO_METRICS_WER_EMBEDDINGS,
+ require_update=False,
+ embedding_transformer=WebEmbeddingTransformer('pl', 'http://localhost:5003', 'fjsd-mkwe-oius-m9h2'),
+ alignment_property_name=WORD_TECHMO_ALIGNMENT_WER_EMBEDDINGS
)
],
experiment_repository=get_repository(),
diff --git a/experiment_data/cached_asr/.gitignore b/experiment_data/cached_asr/.gitignore
index 127221d8df92f47e05599556664cc65c59c7ac14..b8177bca9f52566f1d35f3450716eda975a1af67 100644
--- a/experiment_data/cached_asr/.gitignore
+++ b/experiment_data/cached_asr/.gitignore
@@ -1,3 +1,4 @@
/luna_techmo
/voicelab_cbiz_testset_20220322_techmo
/luna_ajn_polish_asr
+/voicelab_cbiz_testset_20220322_ajn
diff --git a/experiment_data/cached_asr/voicelab_cbiz_testset_20220322_ajn.dvc b/experiment_data/cached_asr/voicelab_cbiz_testset_20220322_ajn.dvc
new file mode 100644
index 0000000000000000000000000000000000000000..107d8c08f496a369d57ef80e2bc54d8c9bdf6446
--- /dev/null
+++ b/experiment_data/cached_asr/voicelab_cbiz_testset_20220322_ajn.dvc
@@ -0,0 +1,5 @@
+outs:
+- md5: 49a38b90f1265a61b90b54f820415011.dir
+ size: 32601414
+ nfiles: 800
+ path: voicelab_cbiz_testset_20220322_ajn
diff --git a/experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/.gitignore b/experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/.gitignore
index 7aff3d637c2b7ca712c0f00c80540e765928aa22..e07b32490887af2a908649169653413934bb654e 100644
--- a/experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/.gitignore
+++ b/experiment_data/pipeline/asr_benchmark_voicelab_cbiz_testset_20220322/.gitignore
@@ -6,3 +6,13 @@
/techmo_spacy
/pos_techmo_alignment_wer
/pos_techmo_metrics_wer
+/ajn_polish_asr
+/word_ajn_metrics_wer
+/word_ajn_alignment_wer
+/ajn_spacy
+/pos_ajn_alignment_wer
+/pos_ajn_metrics_wer
+/word_techmo_metrics_wer_embeddings
+/word_techmo_alignment_wer_embeddings
+/word_ajn_metrics_wer_embeddings
+/word_ajn_alignment_wer_embeddings
diff --git a/run_repro_in_background.sh b/run_repro_in_background.sh
new file mode 100755
index 0000000000000000000000000000000000000000..35d3286fae10f5d9711f80c3293115784f15db59
--- /dev/null
+++ b/run_repro_in_background.sh
@@ -0,0 +1,2 @@
+#!/bin/bash
+dvc repro > repro.log 2>&1 &
diff --git a/sziszapangma/core/transformer/__pycache__/cached_embedding_transformer.cpython-38.pyc b/sziszapangma/core/transformer/__pycache__/cached_embedding_transformer.cpython-38.pyc
index 0ce5df23cfc1255b9563c274caccbf4fb6fc0826..e69f01a8dbd67b79d3cb0386b27827ac6d3021b2 100644
Binary files a/sziszapangma/core/transformer/__pycache__/cached_embedding_transformer.cpython-38.pyc and b/sziszapangma/core/transformer/__pycache__/cached_embedding_transformer.cpython-38.pyc differ
diff --git a/sziszapangma/integration/__pycache__/asr_processor.cpython-38.pyc b/sziszapangma/integration/__pycache__/asr_processor.cpython-38.pyc
index 79aba6f4c6f20a2b241734f0e21c01cd2656659d..adb286664a8b35871a52844a1c97afdc946c595d 100644
Binary files a/sziszapangma/integration/__pycache__/asr_processor.cpython-38.pyc and b/sziszapangma/integration/__pycache__/asr_processor.cpython-38.pyc differ
diff --git a/sziszapangma/integration/asr_processor.py b/sziszapangma/integration/asr_processor.py
index 56a8786b56e78adb5cb2942a2615cd8677be4ea6..4d01c10c7aa94d34378570e739b24b8833a7a9d7 100644
--- a/sziszapangma/integration/asr_processor.py
+++ b/sziszapangma/integration/asr_processor.py
@@ -1,7 +1,7 @@
import json
from abc import ABC, abstractmethod
from pathlib import Path
-from typing import Any, Dict, Optional
+from typing import Any, Dict, Optional, List
import requests
@@ -59,4 +59,28 @@ class AsrPathCacheClient(AsrProcessor):
def call_recognise(self, file_path: str) -> Dict[str, Any]:
path = Path(self._cache_path).joinpath(f"{self.path_to_id[file_path]}.json")
+ print(f'cache path {path}')
+ if not path.exists():
+ raise Exception(f"path not exists {path}")
return json.load(open(path, "r"))
+
+
+class MultipleSourcesAsrProcessor(AsrProcessor):
+ processors: List[AsrProcessor]
+
+ def __init__(self, processors: List[AsrProcessor]):
+ self.processors = processors
+
+ def call_recognise(self, file_path: str) -> Dict[str, Any]:
+ value = None
+ for it in self.processors:
+ if value is None:
+ try:
+ value = it.call_recognise(file_path)
+ except Exception as e:
+ print(e)
+ if value is None:
+ raise Exception("All processors can not process record")
+ else:
+ print(value)
+ return value
diff --git a/sziszapangma/integration/repository/multi_files_experiment_repository.py b/sziszapangma/integration/repository/multi_files_experiment_repository.py
index 238585824d7e071285ce8d89ed9ee93824beb6aa..09394959903bfbeb1845a46d48278b1496086318 100644
--- a/sziszapangma/integration/repository/multi_files_experiment_repository.py
+++ b/sziszapangma/integration/repository/multi_files_experiment_repository.py
@@ -23,9 +23,13 @@ class MultiFilesExperimentRepository(ExperimentRepository):
return self._get_file_path(property_name, record_id).exists()
def update_property_for_key(self, record_id: str, property_name: str, property_value: Any):
+ print(f'update {record_id} {property_name}')
path = self._get_file_path(property_name, record_id)
+ print(path)
path.parent.mkdir(parents=True, exist_ok=True)
+ print('created path')
json.dump(property_value, open(path, "w"))
+ print('value saved')
def delete_property_for_key(self, record_id: str, property_name: str):
self._get_file_path(property_name, record_id).unlink()
diff --git a/sziszapangma/integration/task/__pycache__/asr_task.cpython-38.pyc b/sziszapangma/integration/task/__pycache__/asr_task.cpython-38.pyc
index a34960b88f1c6e4bbe7fb65ec4b80fc16a9658e7..2ec6d5594835708ca2244cb10668efea26f93dc6 100644
Binary files a/sziszapangma/integration/task/__pycache__/asr_task.cpython-38.pyc and b/sziszapangma/integration/task/__pycache__/asr_task.cpython-38.pyc differ
diff --git a/sziszapangma/integration/task/__pycache__/embedding_wer_metrics_task.cpython-38.pyc b/sziszapangma/integration/task/__pycache__/embedding_wer_metrics_task.cpython-38.pyc
index 4918ee5191cb5b9f9f58c552c124e883bb704be7..388f4fe48dd157ff91257c305c0a27b700e0f078 100644
Binary files a/sziszapangma/integration/task/__pycache__/embedding_wer_metrics_task.cpython-38.pyc and b/sziszapangma/integration/task/__pycache__/embedding_wer_metrics_task.cpython-38.pyc differ
diff --git a/sziszapangma/integration/task/asr_task.py b/sziszapangma/integration/task/asr_task.py
index bbb60b2e5aa98a4156d7ca61100ff531624e5613..3c8f444426c7565037603c02c5e3f8d8d1201b8b 100644
--- a/sziszapangma/integration/task/asr_task.py
+++ b/sziszapangma/integration/task/asr_task.py
@@ -36,8 +36,11 @@ class AsrTask(ProcessingTask):
relation_manager: RelationManager,
) -> None:
file_record_path = self._record_path_provider.get_path(record_id)
+ print('before call_recognise', flush=True)
asr_result = self._asr_processor.call_recognise(file_record_path)
+ print('after call_recognise', flush=True)
asr_result["transcription"] = [create_new_word(it) for it in asr_result["transcription"]]
+ print('after create_new_word', flush=True)
experiment_repository.update_property_for_key(
record_id, self._asr_property_name, asr_result
)