refactor

tts/text/pipeline_components/input_parsing/parser.py

@@ -0,0 +1,56 @@
+"""
+Parse raw text sequences into Arpabet and Grapheme sequences so that they can be 
+treated independently.
+
+Inspired by and including code from original Tacotron2 by Keith Ito.
+"""
+
+# pyright: strict
+
+import re
+from typing import List, Union, Any
+
+# Regular expression matching text enclosed in curly braces:
+_curly_re = re.compile(r'(.*?)\{(.+?)\}(.*)')
+
+class ArpabetSequence:
+    def __init__(self, sequence: str):
+        self.sequence = sequence
+    def __eq__(self, other: Any) -> bool:
+        return isinstance(other, ArpabetSequence) and self.sequence == other.sequence
+
+class GraphemeSequence:
+    def __init__(self, sequence: str):
+        self.sequence = sequence
+    def __eq__(self, other: Any) -> bool:
+        return isinstance(other, GraphemeSequence) and self.sequence == other.sequence
+
+ParsedSequence = List[Union[ArpabetSequence, GraphemeSequence]]
+
+def parse_grapheme_and_arpabet_sequence(input_text: str) -> ParsedSequence:
+    """
+    Parse a string into a list of ArpabetSequence and GraphemeSequence so that each 
+    subsequence can be treated independently.
+    """
+    sequence : ParsedSequence = []
+
+    # Check for curly braces and treat their contents as ARPAbet:
+    while len(input_text):
+        input_text = input_text.strip()
+        m = _curly_re.match(input_text)
+        if not m:
+            sequence.append(GraphemeSequence(input_text))
+            break
+
+        before = m.group(1).strip()
+        bracketed = m.group(2).strip()
+        after = m.group(3).strip()
+
+        if before:
+            sequence.append(GraphemeSequence(before))
+        if bracketed:
+            sequence.append(ArpabetSequence(bracketed))
+
+        input_text = after
+
+    return sequence

tts/text/pipeline_components/input_parsing/test_parser.py

@@ -0,0 +1,45 @@
+from tts.text.pipeline_components.input_parsing.parser import ArpabetSequence, GraphemeSequence, parse_grapheme_and_arpabet_sequence
+
+class TestParser:
+
+    def test_only_text(self):
+        parsed = parse_grapheme_and_arpabet_sequence("This is a test")
+        assert len(parsed) == 1
+        assert isinstance(parsed[0], GraphemeSequence)
+
+    def test_only_arpabet(self):
+        parsed = parse_grapheme_and_arpabet_sequence("{T EH1 S T IH0 NG}")
+        assert len(parsed) == 1
+        assert isinstance(parsed[0], ArpabetSequence)
+
+    def test_text_and_arpabet(self):
+        parsed = parse_grapheme_and_arpabet_sequence("This is a {T EH1 S T}")
+        assert len(parsed) == 2
+        assert isinstance(parsed[0], GraphemeSequence)
+        assert isinstance(parsed[1], ArpabetSequence)
+
+    def test_text_and_arpabet_interleved_1(self):
+        parsed = parse_grapheme_and_arpabet_sequence("{DH IH1 S} is a {T EH1 S T}")
+        assert len(parsed) == 3
+        assert isinstance(parsed[0], ArpabetSequence)
+        assert isinstance(parsed[1], GraphemeSequence)
+        assert isinstance(parsed[2], ArpabetSequence)
+
+    def test_text_and_arpabet_interleved_2(self):
+        parsed = parse_grapheme_and_arpabet_sequence("Check out {DH IH1 S} {T EH1 S T} of the {S IH1 S T AH0 M}")
+        assert len(parsed) == 5
+        assert isinstance(parsed[0], GraphemeSequence)
+        assert isinstance(parsed[1], ArpabetSequence)
+        assert isinstance(parsed[2], ArpabetSequence)
+        assert isinstance(parsed[3], GraphemeSequence)
+        assert isinstance(parsed[4], ArpabetSequence)
+
+    def test_text_and_arpabet_interleved_2(self):
+        parsed = parse_grapheme_and_arpabet_sequence("Check out {DH IH1 S} {T EH1 S T} of the {S IH1 S T AH0 M}.")
+        assert len(parsed) == 6
+        assert isinstance(parsed[0], GraphemeSequence)
+        assert isinstance(parsed[1], ArpabetSequence)
+        assert isinstance(parsed[2], ArpabetSequence)
+        assert isinstance(parsed[3], GraphemeSequence)
+        assert isinstance(parsed[4], ArpabetSequence)
+        assert isinstance(parsed[5], GraphemeSequence)