Enabling char_level and compute_CER for aishell recipe (#1554)

* init fix Co-authored-by: Fangjun Kuang <csukuangfj@gmail.com>
k2-fsa · Mar 18, 2024 · bf2f943 · bf2f943
1 parent 2dfd5db
commit bf2f943
Show file tree

Hide file tree

Showing 13 changed files with 80 additions and 26 deletions.
diff --git a/egs/aishell/ASR/conformer_ctc/decode.py b/egs/aishell/ASR/conformer_ctc/decode.py
@@ -419,7 +419,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.exp_dir / f"recogs-{test_set_name}-{key}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         if enable_log:
             logging.info(f"The transcripts are stored in {recog_path}")
 
@@ -432,7 +432,11 @@ def save_results(
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=enable_log
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=enable_log,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/conformer_mmi/decode.py b/egs/aishell/ASR/conformer_mmi/decode.py
@@ -431,7 +431,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.exp_dir / f"recogs-{test_set_name}-{key}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         if enable_log:
             logging.info(f"The transcripts are stored in {recog_path}")
 
@@ -444,7 +444,11 @@ def save_results(
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=enable_log
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=enable_log,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/pruned_transducer_stateless2/decode.py b/egs/aishell/ASR/pruned_transducer_stateless2/decode.py
@@ -390,7 +390,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
@@ -402,7 +402,11 @@ def save_results(
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/pruned_transducer_stateless3/decode.py b/egs/aishell/ASR/pruned_transducer_stateless3/decode.py
@@ -526,7 +526,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
@@ -538,7 +538,11 @@ def save_results(
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/pruned_transducer_stateless7/decode.py b/egs/aishell/ASR/pruned_transducer_stateless7/decode.py
@@ -444,15 +444,19 @@ def save_results(
         for res in results:
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
 
-        store_transcripts(filename=recog_path, texts=results_char)
+        store_transcripts(filename=recog_path, texts=results_char, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
         # ref/hyp pairs.
         errs_filename = params.res_dir / f"errs-{test_set_name}-{params.suffix}.txt"
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/decode.py b/egs/aishell/ASR/pruned_transducer_stateless7_bbpe/decode.py
@@ -581,7 +581,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
@@ -594,7 +594,11 @@ def save_results(
 
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/pruned_transducer_stateless7_streaming/decode.py b/egs/aishell/ASR/pruned_transducer_stateless7_streaming/decode.py
@@ -492,15 +492,19 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
         # ref/hyp pairs.
         errs_filename = params.res_dir / f"errs-{test_set_name}-{params.suffix}.txt"
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/tdnn_lstm_ctc/decode.py b/egs/aishell/ASR/tdnn_lstm_ctc/decode.py
@@ -278,7 +278,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.exp_dir / f"recogs-{test_set_name}-{key}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
@@ -289,7 +289,13 @@ def save_results(
         for res in results:
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
-            wer = write_error_stats(f, f"{test_set_name}-{key}", results_char)
+            wer = write_error_stats(
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=True,
+                compute_CER=True,
+            )
             test_set_wers[key] = wer
 
         logging.info("Wrote detailed error stats to {}".format(errs_filename))

diff --git a/egs/aishell/ASR/transducer_stateless/decode.py b/egs/aishell/ASR/transducer_stateless/decode.py
@@ -327,7 +327,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
 
         # The following prints out WERs, per-word error statistics and aligned
         # ref/hyp pairs.
@@ -338,7 +338,11 @@ def save_results(
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/transducer_stateless_modified-2/decode.py b/egs/aishell/ASR/transducer_stateless_modified-2/decode.py
@@ -372,7 +372,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
@@ -384,7 +384,11 @@ def save_results(
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/transducer_stateless_modified/decode.py b/egs/aishell/ASR/transducer_stateless_modified/decode.py
@@ -376,7 +376,7 @@ def save_results(
     for key, results in results_dict.items():
         recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
@@ -388,7 +388,11 @@ def save_results(
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/whisper/decode.py b/egs/aishell/ASR/whisper/decode.py
@@ -358,7 +358,7 @@ def save_results(
             params.exp_dir / f"recogs-{test_set_name}-{key}-{params.suffix}.txt"
         )
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         if enable_log:
             logging.info(f"The transcripts are stored in {recog_path}")
 
@@ -373,7 +373,11 @@ def save_results(
             results_char.append((res[0], list("".join(res[1])), list("".join(res[2]))))
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results_char, enable_log=enable_log
+                f,
+                f"{test_set_name}-{key}",
+                results_char,
+                enable_log=enable_log,
+                compute_CER=True,
             )
             test_set_wers[key] = wer
 

diff --git a/egs/aishell/ASR/zipformer/decode.py b/egs/aishell/ASR/zipformer/decode.py
@@ -560,7 +560,7 @@ def save_results(
             params.res_dir / f"recogs-{test_set_name}-{key}-{params.suffix}.txt"
         )
         results = sorted(results)
-        store_transcripts(filename=recog_path, texts=results)
+        store_transcripts(filename=recog_path, texts=results, char_level=True)
         logging.info(f"The transcripts are stored in {recog_path}")
 
         # The following prints out WERs, per-word error statistics and aligned
@@ -570,7 +570,11 @@ def save_results(
         )
         with open(errs_filename, "w") as f:
             wer = write_error_stats(
-                f, f"{test_set_name}-{key}", results, enable_log=True
+                f,
+                f"{test_set_name}-{key}",
+                results,
+                enable_log=True,
+                compute_CER=True,
             )
             test_set_wers[key] = wer