Fix lint

experiment/voicelab/voicelab_pipeline.py

+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_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(experiment_repository: ExperimentRepository):
+    record_provider = VoicelabTelcoRecordProvider(
+        ExtensionPathFilter(
+            root_directory=DATASET_DIRECTORY,
+            extension='wav'
+        ),
+        relation_manager_root_path=RELATION_MANAGER_ROOT_PATH
+    )
+    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=experiment_repository,
+        relation_manager_provider=record_provider
+    )
+    experiment_processor.process()
+
+
+def example_run():
+    experiment_repository = MultiFilesExperimentRepository(
+        PIPELINE_DATA_DIRECTORY, EXPERIMENT_NAME)
+    run_voicelab_experiment(experiment_repository)
+
+
+if __name__ == '__main__':
+    example_run()

experiment/voicelab/voicelab_telco_record_provider.py

+from pathlib import Path
+from typing import Dict, Set
+
+from sziszapangma.integration.path_filter import PathFilter
+from sziszapangma.integration.record_id_iterator import RecordIdIterator
+from sziszapangma.integration.record_path_provider import RecordPathProvider
+from sziszapangma.integration.relation_manager_provider import RelationManagerProvider
+from sziszapangma.model.relation_manager import RelationManager, FileRelationManager
+
+
+class VoicelabTelcoRecordProvider(RecordIdIterator, RecordPathProvider, RelationManagerProvider):
+    _path_by_id: Dict[str, str]
+    _relation_manager_root_path: str
+
+    def __init__(self, path_filter: PathFilter, relation_manager_root_path: str):
+        self._path_by_id = dict({
+            self._get_id(it): it
+            for it in path_filter.get_list_of_files()
+        })
+        self._relation_manager_root_path = relation_manager_root_path
+
+    def get_all_records(self) -> Set[str]:
+        return set(self._path_by_id.keys())
+
+    def get_path(self, record_id: str) -> str:
+        return self._path_by_id[record_id]
+
+    def get_relation_manager(self, record_id: str) -> RelationManager:
+        record_path = Path(self._relation_manager_root_path).joinpath(record_id)
+        record_path.mkdir(parents=True, exist_ok=True)
+        return FileRelationManager(
+            str(record_path.joinpath('ab_relations.csv')),
+            str(record_path.joinpath('ab_items.json'))
+        )
+
+    @staticmethod
+    def _get_id(record_file_path: str) -> str:
+        path = record_file_path.replace('.wav', '')
+        return '__'.join(path.split('/')[-2:])

experiment_data/cached_asr/.gitignore

 /luna_techmo
+/voicelab_cbiz_testset_20220322_techmo

experiment_data/cached_asr/voicelab_cbiz_testset_20220322_techmo.dvc

+outs:
+- md5: 94b1709c05bd09b77c5a6850e2f2f373.dir
+  size: 34654307
+  nfiles: 800
+  path: voicelab_cbiz_testset_20220322_techmo

experiment_data/dataset/.gitignore

 /LUNA.PL
+/voicelab_cbiz_testset_20220322

experiment_data/dataset/voicelab_cbiz_testset_20220322.dvc

+outs:
+- md5: 3c2b18e1f1f89e4c5ad7b254e472b25e.dir
+  size: 4803739404
+  nfiles: 1600
+  path: voicelab_cbiz_testset_20220322

mypy.ini

@@ -4,6 +4,9 @@ ignore_missing_imports = True
 [mypy-tensorflow.*]
 ignore_missing_imports = True
 
+[mypy-minio.*]
+ignore_missing_imports = True
+
 [mypy-srsly.*]
 ignore_missing_imports = True
 

sziszapangma/integration/asr_processor.py

@@ -37,7 +37,7 @@ class AsrWebClient(AsrProcessor):
         )
         res = requests.post(self._url, files=files, headers=headers, timeout=600)
         json_response = res.json()
-        print(f'asr processing result {json_response}')
+        print(f"asr processing result {json_response}")
         return json_response
 
 
@@ -45,11 +45,18 @@ class AsrPathCacheClient(AsrProcessor):
     cache_path: str
     path_to_id: Dict[str, str]
 
-    def __init__(self, cache_path: str, record_iterator: RecordIdIterator, record_path_provider: RecordPathProvider):
+    def __init__(
+        self,
+        cache_path: str,
+        record_iterator: RecordIdIterator,
+        record_path_provider: RecordPathProvider,
+    ):
         super(AsrPathCacheClient, self).__init__()
         self._cache_path = cache_path
-        self.path_to_id = {record_path_provider.get_path(it): it for it in record_iterator.get_all_records()}
+        self.path_to_id = {
+            record_path_provider.get_path(it): it for it in record_iterator.get_all_records()
+        }
 
     def call_recognise(self, file_path: str) -> Dict[str, Any]:
-        path = Path(self._cache_path).joinpath(f'{self.path_to_id[file_path]}.json')
-        return json.load(open(path, 'r'))
+        path = Path(self._cache_path).joinpath(f"{self.path_to_id[file_path]}.json")
+        return json.load(open(path, "r"))

sziszapangma/integration/experiment_manager.py

@@ -18,7 +18,7 @@ class ExperimentManager:
         experiment_repository: ExperimentRepository,
         record_id_iterator: RecordIdIterator,
         processing_tasks: List[ProcessingTask],
-        relation_manager_provider: RelationManagerProvider
+        relation_manager_provider: RelationManagerProvider,
     ):
         self._experiment_repository = experiment_repository
         self._record_id_iterator = record_id_iterator

sziszapangma/integration/path_filter.py

@@ -34,6 +34,6 @@ class ExtensionPathFilter(PathFilter):
         """
         Implementation of searching files with extension.
         """
-        path_generator = Path(self._root_directory).glob(f"LUNA.PL/**/*.{self._extension}")
+        path_generator = Path(self._root_directory).glob(f"**/*.{self._extension}")
         all_files = [str(it) for it in path_generator]
         return all_files if self._files_limit is None else all_files[: self._files_limit]

sziszapangma/integration/repository/file_experiment_repository.py

@@ -2,8 +2,6 @@ import json
 import os
 from typing import Any, Dict, Optional, Set
 
-import pandas as pd
-
 from sziszapangma.integration.repository.experiment_repository import ExperimentRepository
 
 

sziszapangma/integration/repository/minio_experiment_repository.py

 import io
 import json
-from typing import Any, Optional, Set
+from typing import Any, List, Optional, Set
 
 from minio import Minio
 
@@ -28,10 +28,11 @@ class MinioExperimentRepository(ExperimentRepository):
 
     def update_property_for_key(self, record_id: str, property_name: str, property_value: Any):
         path = self._get_file_path(property_name, record_id)
-        content_bytes = json.dumps(property_value).encode('utf-8')
+        content_bytes = json.dumps(property_value).encode("utf-8")
         print(self._bucket_name, path)
-        self._client.put_object(self._bucket_name, path, io.BytesIO(content_bytes),
-                                len(content_bytes))
+        self._client.put_object(
+            self._bucket_name, path, io.BytesIO(content_bytes), len(content_bytes)
+        )
 
     def delete_property_for_key(self, record_id: str, property_name: str):
         path = self._get_file_path(property_name, record_id)
@@ -40,24 +41,32 @@ class MinioExperimentRepository(ExperimentRepository):
     def get_property_for_key(self, record_id: str, property_name: str) -> Optional[Any]:
         if self.property_exists(record_id, property_name):
             path = self._get_file_path(property_name, record_id)
-            json_content = self._client.get_object(self._bucket_name, path).read().decode('utf-8')
+            json_content = self._client.get_object(self._bucket_name, path).read().decode("utf-8")
             return json.loads(json_content)
         else:
             return None
 
     def get_all_record_ids(self) -> Set[str]:
-        ids = []
+        ids: List[str] = []
         for property_name in self.get_all_properties():
-            path = f'{self._root_path}{self._experiment_name}/{property_name}/'
-            property_ids = set([obj.object_name.split('/')[-1].replace('.json', '') for obj in
-                                self._client.list_objects(self._bucket_name, path)])
+            path = f"{self._root_path}{self._experiment_name}/{property_name}/"
+            property_ids = set(
+                [
+                    obj.object_name.split("/")[-1].replace(".json", "")
+                    for obj in self._client.list_objects(self._bucket_name, path)
+                ]
+            )
             ids.extend(property_ids)
         return set(ids)
 
     def get_all_properties(self) -> Set[str]:
-        experiment_path = f'{self._root_path}{self._experiment_name}/'
-        return set([obj.object_name.split('/')[-1] for obj in
-                    self._client.list_objects(self._bucket_name, experiment_path)])
+        experiment_path = f"{self._root_path}{self._experiment_name}/"
+        return set(
+            [
+                obj.object_name.split("/")[-1]
+                for obj in self._client.list_objects(self._bucket_name, experiment_path)
+            ]
+        )
 
     def _get_file_path(self, property_name: str, record_id: str) -> str:
-        return f'{self._root_path}{self._experiment_name}/{property_name}/{record_id}.json'
+        return f"{self._root_path}{self._experiment_name}/{property_name}/{record_id}.json"

sziszapangma/integration/repository/multi_files_experiment_repository.py

 import json
 from pathlib import Path
-from typing import Any, Optional, Set
+from typing import Any, List, Optional, Set
 
 from sziszapangma.integration.repository.experiment_repository import ExperimentRepository
 
@@ -24,7 +24,7 @@ class MultiFilesExperimentRepository(ExperimentRepository):
     def update_property_for_key(self, record_id: str, property_name: str, property_value: Any):
         path = self._get_file_path(property_name, record_id)
         path.parent.mkdir(parents=True, exist_ok=True)
-        json.dump(property_value, open(path, 'w'))
+        json.dump(property_value, open(path, "w"))
 
     def delete_property_for_key(self, record_id: str, property_name: str):
         self._get_file_path(property_name, record_id).unlink()
@@ -32,16 +32,17 @@ class MultiFilesExperimentRepository(ExperimentRepository):
     def get_property_for_key(self, record_id: str, property_name: str) -> Optional[Any]:
         if self.property_exists(record_id, property_name):
             path = self._get_file_path(property_name, record_id)
-            return json.load(open(path, 'r'))
+            return json.load(open(path, "r"))
         else:
             return None
 
     def get_all_record_ids(self) -> Set[str]:
-        ids = []
+        ids: List[str] = []
         for property_name in self.get_all_properties():
-            path = Path(self._root_directory).joinpath(self._experiment_name)\
-                .joinpath(property_name)
-            property_ids = set([children.name.replace('.json', '') for children in path.iterdir()])
+            path = (
+                Path(self._root_directory).joinpath(self._experiment_name).joinpath(property_name)
+            )
+            property_ids = set([children.name.replace(".json", "") for children in path.iterdir()])
             ids.extend(property_ids)
         return set(ids)
 
@@ -49,13 +50,10 @@ class MultiFilesExperimentRepository(ExperimentRepository):
         experiment_path = Path(self._root_directory).joinpath(self._experiment_name)
         return set([it.name for it in experiment_path.iterdir()])
 
-
     def _get_file_path(self, property_name: str, record_id: str) -> Path:
-        return Path(self._root_directory) \
-            .joinpath(self._experiment_name) \
-            .joinpath(property_name) \
-            .joinpath(f'{record_id}.json')
-
-
-if __name__ == '__main__':
-    print(list(Path('./').iterdir()))
+        return (
+            Path(self._root_directory)
+            .joinpath(self._experiment_name)
+            .joinpath(property_name)
+            .joinpath(f"{record_id}.json")
+        )

sziszapangma/integration/repository/multiple_experiment_repository.py

-from typing import Any, Optional, Set, List
+from typing import Any, List, Optional, Set
 
 from sziszapangma.integration.repository.experiment_repository import ExperimentRepository
 

sziszapangma/integration/task/gold_transcript_task.py

@@ -27,7 +27,12 @@ class GoldTranscriptTask(ProcessingTask):
             is not None
         )
 
-    def run_single_process(self, record_id: str, experiment_repository: ExperimentRepository, relation_manager: RelationManager,):
+    def run_single_process(
+        self,
+        record_id: str,
+        experiment_repository: ExperimentRepository,
+        relation_manager: RelationManager,
+    ):
         experiment_repository.update_property_for_key(
             record_id,
             self._gold_transcript_property_name,

sziszapangma/model/relation_manager.py

@@ -114,4 +114,3 @@ class FileRelationManager(RelationManager):
     def clear_all(self) -> None:
         self.items_dict.clear()
         self.relations_dataframe = self.relations_dataframe[0:0]
-