Merge pull request #93 from alibaba-damo-academy/dev_lzr

update paraformer-large model RESULTS.md and support for turning off timestamps

egs_modelscope/asr/paraformer/speech_paraformer-large_asr_nat-zh-cn-16k-aishell1-vocab8404-pytorch/RESULTS.md

@@ -0,0 +1,23 @@
+# Paraformer-Large
+- Model link: <https://www.modelscope.cn/models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-aishell1-vocab8404-pytorch/summary>
+- Model size: 220M
+
+# Environments
+- date: `Fri Feb 10 13:34:24 CST 2023`
+- python version: `3.7.12`
+- FunASR version: `0.1.6`
+- pytorch version: `pytorch 1.7.0`
+- Git hash: ``
+- Commit date: ``
+
+# Beachmark Results
+
+## AISHELL-1
+- Decode config:
+  - Decode without CTC
+  - Decode without LM
+
+| testset CER(%) | base model|finetune model |
+|:--------------:|:---------:|:-------------:|
+| dev            | 1.75      |1.62           |
+| test           | 1.95      |1.78           |

egs_modelscope/asr/paraformer/speech_paraformer-large_asr_nat-zh-cn-16k-aishell2-vocab8404-pytorch/RESULTS.md

@@ -0,0 +1,25 @@
+# Paraformer-Large
+- Model link: <https://www.modelscope.cn/models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-aishell2-vocab8404-pytorch/summary>
+- Model size: 220M
+
+# Environments
+- date: `Fri Feb 10 13:34:24 CST 2023`
+- python version: `3.7.12`
+- FunASR version: `0.1.6`
+- pytorch version: `pytorch 1.7.0`
+- Git hash: ``
+- Commit date: ``
+
+# Beachmark Results
+
+## AISHELL-2
+- Decode config: 
+  - Decode without CTC
+  - Decode without LM
+
+| testset      | base model|finetune model|
+|:------------:|:---------:|:------------:|
+| dev_ios      | 2.80      |2.60          |
+| test_android | 3.13      |2.84          |
+| test_ios     | 2.85      |2.82          |
+| test_mic     | 3.06      |2.88          |

egs_modelscope/asr/paraformer/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/RESULTS.md

@@ -0,0 +1,75 @@
+# Paraformer-Large
+- Model link: <https://www.modelscope.cn/models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch/summary>
+- Model size: 220M
+
+# Environments
+- date: `Tue Nov 22 18:48:39 CST 2022`
+- python version: `3.7.12`
+- FunASR version: `0.1.0`
+- pytorch version: `pytorch 1.7.0`
+- Git hash: ``
+- Commit date: ``
+
+# Beachmark Results
+
+## AISHELL-1
+- Decode config: 
+  - Decode without CTC
+  - Decode without LM
+
+| testset   | CER(%)|
+|:---------:|:-----:|
+| dev       | 1.75  |
+| test      | 1.95  |
+
+## AISHELL-2
+- Decode config: 
+  - Decode without CTC
+  - Decode without LM
+
+| testset      | CER(%)|
+|:------------:|:-----:|
+| dev_ios      | 2.80  |
+| test_android | 3.13  |
+| test_ios     | 2.85  |
+| test_mic     | 3.06  |
+
+## Wenetspeech
+- Decode config: 
+  - Decode without CTC
+  - Decode without LM
+
+| testset   | CER(%)|
+|:---------:|:-----:|
+| dev       | 3.57  |
+| test      | 6.97  |
+| test_net  | 6.74  |
+
+## SpeechIO TIOBE
+- Decode config 1:
+  - Decode without CTC
+  - Decode without LM
+  - With text norm
+- Decode config 2:
+  - Decode without CTC
+  - Decode with Transformer-LM
+  - LM weight: 0.15
+  - With text norm
+
+| testset | w/o LM | w/ LM |
+|:------------------:|:----:|:----:|
+|SPEECHIO_ASR_ZH00001| 0.49 | 0.35 |
+|SPEECHIO_ASR_ZH00002| 3.23 | 2.86 |
+|SPEECHIO_ASR_ZH00003| 1.13 | 0.80 |
+|SPEECHIO_ASR_ZH00004| 1.33 | 1.10 |
+|SPEECHIO_ASR_ZH00005| 1.41 | 1.18 |
+|SPEECHIO_ASR_ZH00006| 5.25 | 4.85 |
+|SPEECHIO_ASR_ZH00007| 5.51 | 4.97 |
+|SPEECHIO_ASR_ZH00008| 3.69 | 3.18 |
+|SPEECHIO_ASR_ZH00009| 3.02 | 2.78 |
+|SPEECHIO_ASR_ZH000010| 3.35 | 2.99 |
+|SPEECHIO_ASR_ZH000011| 1.54 | 1.25 |
+|SPEECHIO_ASR_ZH000012| 2.06 | 1.68 |
+|SPEECHIO_ASR_ZH000013| 2.57 | 2.25 |
+|SPEECHIO_ASR_ZH000014| 3.86 | 3.08 |
+|SPEECHIO_ASR_ZH000015| 3.34 | 2.67 |

funasr/bin/asr_inference.py

@@ -453,7 +453,7 @@ def inference_modelscope(
                     ibest_writer["score"][key] = str(hyp.score)
                 
                 if text is not None:
-                    text_postprocessed = postprocess_utils.sentence_postprocess(token)
+                    text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                     item = {'key': key, 'value': text_postprocessed}
                     asr_result_list.append(item)
                     finish_count += 1

funasr/bin/asr_inference_paraformer.py

@@ -428,7 +428,11 @@ def inference_modelscope(
         format="%(asctime)s (%(module)s:%(lineno)d) %(levelname)s: %(message)s",
     )
 
-    hotword_list_or_file = param_dict['hotword']
+    if param_dict is not None:
+        hotword_list_or_file = param_dict.get('hotword')
+    else:
+        hotword_list_or_file = None
+
     if ngpu >= 1 and torch.cuda.is_available():
         device = "cuda"
     else:
@@ -539,7 +543,7 @@ def inference_modelscope(
                         ibest_writer["rtf"][key] = rtf_cur
 
                     if text is not None:
-                        text_postprocessed = postprocess_utils.sentence_postprocess(token)
+                        text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                         item = {'key': key, 'value': text_postprocessed}
                         asr_result_list.append(item)
                         finish_count += 1

funasr/bin/asr_inference_paraformer_timestamp.py

@@ -436,7 +436,7 @@ def inference(
                     ibest_writer["score"][key] = str(hyp.score)
     
                 if text is not None:
-                    text_postprocessed = postprocess_utils.sentence_postprocess(token)
+                    text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                     item = {'key': key, 'value': text_postprocessed}
                     asr_result_list.append(item)
                     finish_count += 1

funasr/bin/asr_inference_paraformer_vad.py

@@ -241,6 +241,11 @@ def inference_modelscope(
             allow_variable_data_keys=allow_variable_data_keys,
             inference=True,
         )
+
+        if param_dict is not None:
+            use_timestamp = param_dict.get('use_timestamp', True)
+        else:
+            use_timestamp = True
         
         finish_count = 0
         file_count = 1
@@ -284,8 +289,10 @@ def inference_modelscope(
                 text, token, token_int = result[0], result[1], result[2]
                 time_stamp = None if len(result) < 4 else result[3]
                
-                
-                postprocessed_result = postprocess_utils.sentence_postprocess(token, time_stamp)
+                if use_timestamp and time_stamp is not None:
+                    postprocessed_result = postprocess_utils.sentence_postprocess(token, time_stamp)
+                else:
+                    postprocessed_result = postprocess_utils.sentence_postprocess(token)
                 text_postprocessed = ""
                 time_stamp_postprocessed = ""
                 text_postprocessed_punc = postprocessed_result
@@ -293,9 +300,11 @@ def inference_modelscope(
                     text_postprocessed, time_stamp_postprocessed, word_lists = postprocessed_result[0], \
                                                                                postprocessed_result[1], \
                                                                                postprocessed_result[2]
-                    text_postprocessed_punc = text_postprocessed
-                    if len(word_lists) > 0 and text2punc is not None:
-                        text_postprocessed_punc, punc_id_list = text2punc(word_lists, 20)
+                else:
+                    text_postprocessed, word_lists = postprocessed_result[0], postprocessed_result[1]
+                text_postprocessed_punc = text_postprocessed
+                if len(word_lists) > 0 and text2punc is not None:
+                    text_postprocessed_punc, punc_id_list = text2punc(word_lists, 20)
 
                 
                 item = {'key': key, 'value': text_postprocessed_punc}

funasr/bin/asr_inference_paraformer_vad_punc.py

@@ -570,6 +570,11 @@ def inference_modelscope(
             allow_variable_data_keys=allow_variable_data_keys,
             inference=True,
         )
+
+        if param_dict is not None:
+            use_timestamp = param_dict.get('use_timestamp', True)
+        else:
+            use_timestamp = True
     
         finish_count = 0
         file_count = 1
@@ -612,8 +617,11 @@ def inference_modelscope(
                 result = result_segments[0]
                 text, token, token_int = result[0], result[1], result[2]
                 time_stamp = None if len(result) < 4 else result[3]
-    
-                postprocessed_result = postprocess_utils.sentence_postprocess(token, time_stamp)
+   
+                if use_timestamp and time_stamp is not None: 
+                    postprocessed_result = postprocess_utils.sentence_postprocess(token, time_stamp)
+                else:
+                    postprocessed_result = postprocess_utils.sentence_postprocess(token)
                 text_postprocessed = ""
                 time_stamp_postprocessed = ""
                 text_postprocessed_punc = postprocessed_result
@@ -621,9 +629,12 @@ def inference_modelscope(
                     text_postprocessed, time_stamp_postprocessed, word_lists = postprocessed_result[0], \
                                                                                postprocessed_result[1], \
                                                                                postprocessed_result[2]
-                    text_postprocessed_punc = text_postprocessed
-                    if len(word_lists) > 0 and text2punc is not None:
-                        text_postprocessed_punc, punc_id_list = text2punc(word_lists, 20)
+                else:
+                    text_postprocessed, word_lists = postprocessed_result[0], postprocessed_result[1]
+
+                text_postprocessed_punc = text_postprocessed
+                if len(word_lists) > 0 and text2punc is not None:
+                    text_postprocessed_punc, punc_id_list = text2punc(word_lists, 20)
     
                 item = {'key': key, 'value': text_postprocessed_punc}
                 if text_postprocessed != "":

funasr/bin/asr_inference_uniasr.py

@@ -492,7 +492,7 @@ def inference_modelscope(
                     ibest_writer["score"][key] = str(hyp.score)
     
                 if text is not None:
-                    text_postprocessed = postprocess_utils.sentence_postprocess(token)
+                    text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                     item = {'key': key, 'value': text_postprocessed}
                     asr_result_list.append(item)
                     finish_count += 1

funasr/bin/asr_inference_uniasr_vad.py

@@ -492,7 +492,7 @@ def inference_modelscope(
                     ibest_writer["score"][key] = str(hyp.score)
     
                 if text is not None:
-                    text_postprocessed = postprocess_utils.sentence_postprocess(token)
+                    text_postprocessed, _ = postprocess_utils.sentence_postprocess(token)
                     item = {'key': key, 'value': text_postprocessed}
                     asr_result_list.append(item)
                     finish_count += 1

funasr/utils/postprocess_utils.py

@@ -232,5 +232,9 @@ def sentence_postprocess(words: List[Any], time_stamp: List[List] = None):
         return sentence, ts_lists, real_word_lists
     else:
         word_lists = abbr_dispose(word_lists)
+        real_word_lists = []
+        for ch in word_lists:
+            if ch != ' ':
+                real_word_lists.append(ch)
         sentence = ''.join(word_lists).strip()
-        return sentence
+        return sentence, real_word_lists