EngineX-Iluvatar/enginex-mr_series-sherpa-onnx
Archived
16
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Add homonphone replacer example for Python API. (#2161)

Browse Source
This commit is contained in:
Fangjun Kuang
2025-04-29 15:59:34 +08:00
committed by GitHub
parent 73dba27145
commit 51f8824219
7 changed files with 133 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
python-api-examples/speech-recognition-from-microphone-with-endpoint-detection-alsa.py
View File

@@ -106,6 +106,27 @@ def get_args():
""",
)
parser.add_argument(
"--hr-dict-dir",
type=str,
default="",
help="If not empty, it is the jieba dict directory for homophone replacer",
)
parser.add_argument(
"--hr-lexicon",
type=str,
default="",
help="If not empty, it is the lexicon.txt for homophone replacer",
)
parser.add_argument(
"--hr-rule-fsts",
type=str,
default="",
help="If not empty, it is the replace.fst for homophone replacer",
)
parser.add_argument(
"--device-name",
type=str,
@@ -159,6 +180,9 @@ def create_recognizer(args):
hotwords_file=args.hotwords_file,
hotwords_score=args.hotwords_score,
blank_penalty=args.blank_penalty,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
return recognizer

24
python-api-examples/speech-recognition-from-microphone-with-endpoint-detection.py
View File

@@ -114,6 +114,27 @@ def get_args():
""",
)
parser.add_argument(
"--hr-dict-dir",
type=str,
default="",
help="If not empty, it is the jieba dict directory for homophone replacer",
)
parser.add_argument(
"--hr-lexicon",
type=str,
default="",
help="If not empty, it is the lexicon.txt for homophone replacer",
)
parser.add_argument(
"--hr-rule-fsts",
type=str,
default="",
help="If not empty, it is the replace.fst for homophone replacer",
)
return parser.parse_args()
@@ -142,6 +163,9 @@ def create_recognizer(args):
hotwords_file=args.hotwords_file,
hotwords_score=args.hotwords_score,
blank_penalty=args.blank_penalty,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
return recognizer

24
python-api-examples/speech-recognition-from-microphone.py
View File

@@ -123,6 +123,27 @@ def get_args():
""",
)
parser.add_argument(
"--hr-dict-dir",
type=str,
default="",
help="If not empty, it is the jieba dict directory for homophone replacer",
)
parser.add_argument(
"--hr-lexicon",
type=str,
default="",
help="If not empty, it is the lexicon.txt for homophone replacer",
)
parser.add_argument(
"--hr-rule-fsts",
type=str,
default="",
help="If not empty, it is the replace.fst for homophone replacer",
)
return parser.parse_args()
@@ -148,6 +169,9 @@ def create_recognizer(args):
hotwords_file=args.hotwords_file,
hotwords_score=args.hotwords_score,
blank_penalty=args.blank_penalty,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
return recognizer

23
python-api-examples/speech-recognition-from-url.py
View File

@@ -110,6 +110,26 @@ def get_args():
""",
)
parser.add_argument(
"--hr-dict-dir",
type=str,
default="",
help="If not empty, it is the jieba dict directory for homophone replacer",
)
parser.add_argument(
"--hr-lexicon",
type=str,
default="",
help="If not empty, it is the lexicon.txt for homophone replacer",
)
parser.add_argument(
"--hr-rule-fsts",
type=str,
default="",
help="If not empty, it is the replace.fst for homophone replacer",
)
return parser.parse_args()
@@ -133,6 +153,9 @@ def create_recognizer(args):
rule3_min_utterance_length=300, # it essentially disables this rule
hotwords_file=args.hotwords_file,
hotwords_score=args.hotwords_score,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
return recognizer

36
python-api-examples/vad-with-non-streaming-asr.py
View File

@@ -273,6 +273,27 @@ def get_args():
help="Feature dimension. Must match the one expected by the model",
)
parser.add_argument(
"--hr-dict-dir",
type=str,
default="",
help="If not empty, it is the jieba dict directory for homophone replacer",
)
parser.add_argument(
"--hr-lexicon",
type=str,
default="",
help="If not empty, it is the lexicon.txt for homophone replacer",
)
parser.add_argument(
"--hr-rule-fsts",
type=str,
default="",
help="If not empty, it is the replace.fst for homophone replacer",
)
return parser.parse_args()
@@ -312,6 +333,9 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
decoding_method=args.decoding_method,
blank_penalty=args.blank_penalty,
debug=args.debug,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
elif args.paraformer:
assert len(args.sense_voice) == 0, args.sense_voice
@@ -334,6 +358,9 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
feature_dim=args.feature_dim,
decoding_method=args.decoding_method,
debug=args.debug,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
elif args.sense_voice:
assert len(args.whisper_encoder) == 0, args.whisper_encoder
@@ -352,6 +379,9 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
num_threads=args.num_threads,
use_itn=True,
debug=args.debug,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
elif args.whisper_encoder:
assert_file_exists(args.whisper_encoder)
@@ -373,6 +403,9 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
language=args.whisper_language,
task=args.whisper_task,
tail_paddings=args.whisper_tail_paddings,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
elif args.moonshine_preprocessor:
assert_file_exists(args.moonshine_preprocessor)
@@ -389,6 +422,9 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
num_threads=args.num_threads,
decoding_method=args.decoding_method,
debug=args.debug,
hr_dict_dir=args.hr_dict_dir,
hr_rule_fsts=args.hr_rule_fsts,
hr_lexicon=args.hr_lexicon,
)
else:
raise ValueError("Please specify at least one model")