Bug fix for MontrealCorpusTools#459 (MontrealCorpusTools#460)

* Bug fix for MontrealCorpusTools#459 * Update tests for new models
Aditya514 · Jun 4, 2022 · 144e121 · 144e121
1 parent 3944163
commit 144e121
Show file tree

Hide file tree

Showing 2 changed files with 9 additions and 9 deletions.
diff --git a/montreal_forced_aligner/textgrid.py b/montreal_forced_aligner/textgrid.py
@@ -179,10 +179,10 @@ def export_textgrid(
         if include_utterance_text:
             for u in data["utterances"]:
                 tg.tierDict[utterance_tier_name].entryList.append(u.to_tg_interval())
-    for tier in tg.tierDict.values():
-        if tier.entryList[-1][1] > tg.maxTimestamp:
-            tier.entryList[-1] = Interval(
-                tier.entryList[-1].start, tg.maxTimestamp, tier.entryList[-1].label
-            )
     if has_data:
+        for tier in tg.tierDict.values():
+            if len(tier.entryList) > 0 and tier.entryList[-1][1] > tg.maxTimestamp:
+                tier.entryList[-1] = Interval(
+                    tier.entryList[-1].start, tg.maxTimestamp, tier.entryList[-1].label
+                )
         tg.save(output_path, includeBlankSpaces=True, format=output_format, reportingMode="error")
diff --git a/tests/test_g2p.py b/tests/test_g2p.py
@@ -35,17 +35,17 @@ def test_scored_match(english_g2p_model, temp_dir):
     g2p_model = G2PModel(english_g2p_model)
     pron_scores = {
         ("read", "R IY1 D"): 12.3903465,
-        ("read", "R EH1 D"): 12.2733212,
+        ("read", "R EH1 D"): 3.03866529,
         ("theatres", "TH IY1 AH0 T ER0 Z"): 12.3603706,
         ("theatres", "DH IY1"): 39.7762604,
         ("the", "DH"): 11.7815981,
         ("the", "TH"): 12.3016987,
         ("the", "DH AH0"): 16.9708042,
         ("the", "DH IY1"): 16.2190933,
         ("the", "DH IY0"): 15.3279858,
-        ("them", "DH EH0 M"): 14.7657909,
+        ("them", "DH EH0 M"): 9.89364433,
         ("them", "DH AH0 M"): 11.9653378,
-        ("them", "DH EY1"): 28.7747478,
+        ("them", "DH EY1"): 25.8712311,
     }
     fst = pynini.Fst.read(g2p_model.fst_path)
     for (word, pron), absolute_reference in pron_scores.items():
@@ -57,7 +57,7 @@ def test_scored_match(english_g2p_model, temp_dir):
             input_token_type="utf8",
             output_token_type=pynini.SymbolTable.read_text(g2p_model.sym_path),
         )
-        assert absolute_score == absolute_reference
+        assert abs(absolute_score - absolute_reference) < 2
 
 
 def test_training(basic_dict_path, basic_g2p_model_path, temp_dir):