fix apostrophe, hyphen

tts/text/pipeline/test_english_pipeline_v1.py

@@ -58,3 +58,8 @@ class TestEnglishPipelineV1:
             11, # Space
             2, # Exclamation mark
         ]
+
+    def test_apostrophe_word(self):
+        pipeline = self.build_pipeline_instance()
+        output = pipeline.user_input_to_sequence("you're") 
+        assert output == [145, 137, 130] # YOU'RE  Y UH1 R   ~   145  137  130

tts/text/pipeline_components/input_parsing/parser.py

@@ -23,6 +23,8 @@ class ArpabetSequence:
         self.sequence = sequence
     def __eq__(self, other: Any) -> bool:
         return isinstance(other, ArpabetSequence) and self.sequence == other.sequence
+    def __repr__(self):
+        return f"<ArpabetSequence: {self.sequence}>"
 
 
 class GraphemeSequence:
@@ -34,6 +36,8 @@ class GraphemeSequence:
         self.sequence = sequence
     def __eq__(self, other: Any) -> bool:
         return isinstance(other, GraphemeSequence) and self.sequence == other.sequence
+    def __repr__(self):
+        return f"<GraphemeSequence: {self.sequence}>"
 
 
 ParsedSequence = List[Union[ArpabetSequence, GraphemeSequence]]

tts/text/pipeline_components/phonetization/base_phonetization.py

@@ -4,7 +4,11 @@ from typing import List
 from text.pipeline_components.phonetization.cmudict import CMUDict
 
 # Regex taken from open source tacotron2 extensions
-_arpabet_split_re = re.compile(r'[^\s]*{.*?}[^\s]*|[\w]+|[^\w\s]+')
+# NB: This was broken and couldn't handle apostrophes
+#_arpabet_split_re = re.compile(r'[^\s]*{.*?}[^\s]*|[\w]+|[^\w\s]+')
+
+# Modified regex to not split apostrophe (eg. "you're", "I'm", etc.) or hyphenation (eg "avant-garde")
+_arpabet_split_re = re.compile(r'[^\s]*{.*?}[^\s]*|[\w]+\'[\w]+|[\w]+\-[\w]+|[\w]+|[^\w\s]+')
 
 class BasePhonetization:
     def __init__(self, phoneme_dictionary: CMUDict):

tts/text/pipeline_components/phonetization/test_english_phonetization.py

@@ -33,6 +33,27 @@ class TestEnglishPhonetization:
             ArpabetSequence("T EH1 S T"),
         ]
 
+    def test_words_with_apostrophe(self):
+        phoneme_dictionary = CMUDict(DEFAULT_CMUDICT_PATH)
+        phonetization = EnglishPhonetization(phoneme_dictionary)
+        output = phonetization.try_convert_to_phonemes("I'm can't you're we're ain't")
+        assert output == [
+            ArpabetSequence("AY1 M"), 
+            ArpabetSequence("K AE1 N T"), 
+            ArpabetSequence("Y UH1 R"),
+            ArpabetSequence("W IY1 R"),
+            ArpabetSequence("EY1 N T"),
+        ]
+
+    def test_words_with_hyphenation(self):
+        phoneme_dictionary = CMUDict(DEFAULT_CMUDICT_PATH)
+        phonetization = EnglishPhonetization(phoneme_dictionary)
+        output = phonetization.try_convert_to_phonemes("avant-garde bake-off")
+        assert output == [
+            ArpabetSequence("AH0 V AA1 N T G AA1 R D"), 
+            ArpabetSequence("B EY1 K AO1 F"), 
+        ]
+
     def test_punctuation(self):
         phoneme_dictionary = CMUDict(DEFAULT_CMUDICT_PATH)
         phonetization = EnglishPhonetization(phoneme_dictionary)