Refactor prep_segments in SVS

egs2/TEMPLATE/asr1/db.sh

@@ -181,6 +181,7 @@ AMEBOSHI=
 ITAKO=
 NATSUME=
 KIRITAN=
+NAMINE=
 
 # For only CMU TIR environment
 if [[ "$(hostname)" == tir* ]]; then

egs2/TEMPLATE/asr1/pyscripts/utils/check_align.py

@@ -2,6 +2,7 @@
 import argparse
 import math
 import os
+import re
 import sys
 
 from espnet2.fileio.read_text import read_label
@@ -61,6 +62,18 @@ def compare(key, score, label):
         pre_phn = phns[-1]
         for p in phns:
             if index >= len(label):
+                pattern = r"_[0]*"
+                key_name = re.split(pattern, key[:-5], 1)[-1]
+                print(
+                    "Error in {}, copy this code to `get_error_dict`"
+                    ' in prep_segments.py under `if input_type == "xml"`.\n'
+                    '"{}": [\n'
+                    "   lambda i, labels, segment, segments, threshold: "
+                    "add_pause(labels, segment, segments, threshold)\n"
+                    '   if (labels[i].lyric == "{}" and labels[i - 1].lyric == "{}")\n'
+                    "   else (labels, segment, segments, False),\n"
+                    "],".format(key, key_name, score[i][2], score[i - 1][2])
+                )
                 raise ValueError("Lyrics are longer than phones in {}".format(key))
             elif label[index][2] == p:
                 index += 1
@@ -71,6 +84,18 @@ def compare(key, score, label):
                     )
                 )
     if index != len(label):
+        pattern = r"_[0]*"
+        key_name = re.split(pattern, key[:-5], 1)[-1]
+        print(
+            "Error in {}, copy this code to `get_error_dict` in prep_segments.py"
+            ' under `if input_type == "hts"`.\n'
+            '"{}": [\n'
+            "   lambda i, labels, segment, segments, threshold: "
+            "add_pause(labels, segment, segments, threshold)\n"
+            '   if (labels[i].label_id == "{}" and labels[i - 1].label_id == "{}")\n'
+            "   else (labels, segment, segments, False),\n"
+            "],".format(key, key_name, label[index][2], label[index - 1][2])
+        )
         raise ValueError("Phones are longer than lyrics in {}.".format(key))
     return score