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

@@ -30,7 +30,7 @@ $EXE \
   --model-type=nemo_transducer \
   --model-type=nemo_transducer \
   ./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/test_wavs/example.wav
   ./sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19/test_wavs/example.wav
 
 
-rm sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19
+rm -rf sherpa-onnx-nemo-transducer-giga-am-v2-russian-2025-04-19
 
 
 
 
 log "------------------------------------------------------------------------"
 log "------------------------------------------------------------------------"

@@ -28,7 +28,7 @@ on:
   workflow_dispatch:
   workflow_dispatch:
 
 
 concurrency:
 concurrency:
-  group: aarch64-linux-gnu-shared-${{ github.ref }}
+  group: rknn-linux-aarch64-${{ github.ref }}
   cancel-in-progress: true
   cancel-in-progress: true
 
 
 jobs:
 jobs:

@@ -107,6 +107,27 @@ def get_args():
     )
     )
 
 
     parser.add_argument(
     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",
         "--device-name",
         type=str,
         type=str,
         required=True,
         required=True,
@@ -159,6 +180,9 @@ def create_recognizer(args):
         hotwords_file=args.hotwords_file,
         hotwords_file=args.hotwords_file,
         hotwords_score=args.hotwords_score,
         hotwords_score=args.hotwords_score,
         blank_penalty=args.blank_penalty,
         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
     return recognizer
 
 

@@ -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()
     return parser.parse_args()
 
 
 
 
@@ -142,6 +163,9 @@ def create_recognizer(args):
         hotwords_file=args.hotwords_file,
         hotwords_file=args.hotwords_file,
         hotwords_score=args.hotwords_score,
         hotwords_score=args.hotwords_score,
         blank_penalty=args.blank_penalty,
         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
     return recognizer
 
 

@@ -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()
     return parser.parse_args()
 
 
 
 
@@ -148,6 +169,9 @@ def create_recognizer(args):
         hotwords_file=args.hotwords_file,
         hotwords_file=args.hotwords_file,
         hotwords_score=args.hotwords_score,
         hotwords_score=args.hotwords_score,
         blank_penalty=args.blank_penalty,
         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
     return recognizer
 
 

@@ -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()
     return parser.parse_args()
 
 
@@ -133,6 +153,9 @@ def create_recognizer(args):
         rule3_min_utterance_length=300,  # it essentially disables this rule
         rule3_min_utterance_length=300,  # it essentially disables this rule
         hotwords_file=args.hotwords_file,
         hotwords_file=args.hotwords_file,
         hotwords_score=args.hotwords_score,
         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
     return recognizer
 
 

@@ -273,6 +273,27 @@ def get_args():
         help="Feature dimension. Must match the one expected by the model",
         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()
     return parser.parse_args()
 
 
 
 
@@ -312,6 +333,9 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
             decoding_method=args.decoding_method,
             decoding_method=args.decoding_method,
             blank_penalty=args.blank_penalty,
             blank_penalty=args.blank_penalty,
             debug=args.debug,
             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:
     elif args.paraformer:
         assert len(args.sense_voice) == 0, args.sense_voice
         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,
             feature_dim=args.feature_dim,
             decoding_method=args.decoding_method,
             decoding_method=args.decoding_method,
             debug=args.debug,
             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:
     elif args.sense_voice:
         assert len(args.whisper_encoder) == 0, args.whisper_encoder
         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,
             num_threads=args.num_threads,
             use_itn=True,
             use_itn=True,
             debug=args.debug,
             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:
     elif args.whisper_encoder:
         assert_file_exists(args.whisper_encoder)
         assert_file_exists(args.whisper_encoder)
@@ -373,6 +403,9 @@ def create_recognizer(args) -> sherpa_onnx.OfflineRecognizer:
             language=args.whisper_language,
             language=args.whisper_language,
             task=args.whisper_task,
             task=args.whisper_task,
             tail_paddings=args.whisper_tail_paddings,
             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:
     elif args.moonshine_preprocessor:
         assert_file_exists(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,
             num_threads=args.num_threads,
             decoding_method=args.decoding_method,
             decoding_method=args.decoding_method,
             debug=args.debug,
             debug=args.debug,
+            hr_dict_dir=args.hr_dict_dir,
+            hr_rule_fsts=args.hr_rule_fsts,
+            hr_lexicon=args.hr_lexicon,
         )
         )
     else:
     else:
         raise ValueError("Please specify at least one model")
         raise ValueError("Please specify at least one model")