phoonnx 0.0.0__py3-none-any.whl

phoonnx/phonemizers/he.py

@@ -0,0 +1,67 @@
+import os.path
+
+import requests
+
+from phoonnx.phonemizers.base import BasePhonemizer
+from phoonnx.config import Alphabet
+
+
+class PhonikudPhonemizer(BasePhonemizer):
+    dl_url = "https://huggingface.co/thewh1teagle/phonikud-onnx/resolve/main/phonikud-1.0.int8.onnx"
+
+    def __init__(self, model: str = None, diacritics=True):
+        from phonikud_onnx import Phonikud
+        from phonikud import phonemize
+        self.g2p = phonemize
+        self.diacritics = diacritics
+        if model is None:
+            base_path = os.path.expanduser("~/.local/share/phonikud")
+            fname = self.dl_url.split("/")[-1]
+            model = f"{base_path}/{fname}"
+            if not os.path.isfile(model):
+                os.makedirs(base_path, exist_ok=True)
+                # TODO - streaming download
+                data = requests.get(self.dl_url).content
+                with open(model, "wb") as f:
+                    f.write(data)
+        self.phonikud = Phonikud(model) if diacritics else None
+        super().__init__(Alphabet.IPA)
+
+    @classmethod
+    def get_lang(cls, target_lang: str) -> str:
+        """
+        Validates and returns the closest supported language code.
+
+        Args:
+            target_lang (str): The language code to validate.
+
+        Returns:
+            str: The validated language code.
+
+        Raises:
+            ValueError: If the language code is unsupported.
+        """
+        # this check is here only to throw an exception if invalid language is provided
+        return cls.match_lang(target_lang, ["he"])
+
+    def phonemize_string(self, text: str, lang: str = "he") -> str:
+        """
+        """
+        lang = self.get_lang(lang)
+        if self.diacritics:
+            text = self.phonikud.add_diacritics(text)
+        return self.g2p(text)
+
+
+if __name__ == "__main__":
+    #text = "מתכת יקרה"
+    text = 'שָׁלוֹם עוֹלָם'
+
+    pho = PhonikudPhonemizer(diacritics=False)
+    lang = "he"
+
+    print(f"\n--- Getting phonemes for '{text}' ---")
+    phonemes = pho.phonemize(text, lang)
+    print(f"  Phonemes: {phonemes}")
+    # --- Getting phonemes for 'שָׁלוֹם עוֹלָם' ---
+    #   Phonemes: [('ʃalˈom ʔolˈam', '.', True)]

phoonnx/phonemizers/ja.py

@@ -0,0 +1,119 @@
+from phoonnx.phonemizers.base import BasePhonemizer
+from phoonnx.config import Alphabet
+
+class OpenJTaklPhonemizer(BasePhonemizer):
+
+    def __init__(self, alphabet=Alphabet.IPA):
+        assert alphabet in [Alphabet.HEPBURN, Alphabet.KANA]
+        import pyopenjtalk
+        self.g2p = pyopenjtalk.g2p
+        super().__init__(alphabet)
+
+    @classmethod
+    def get_lang(cls, target_lang: str) -> str:
+        """
+        Validates and returns the closest supported language code.
+
+        Args:
+            target_lang (str): The language code to validate.
+
+        Returns:
+            str: The validated language code.
+
+        Raises:
+            ValueError: If the language code is unsupported.
+        """
+        # this check is here only to throw an exception if invalid language is provided
+        return cls.match_lang(target_lang, ["ja"])
+
+    def phonemize_string(self, text: str, lang: str = "ja") -> str:
+        """
+        """
+        lang = self.get_lang(lang)
+        return self.g2p(text, kana=self.alphabet == Alphabet.KANA)
+
+
+class CutletPhonemizer(BasePhonemizer):
+
+    def __init__(self, alphabet=Alphabet.HEPBURN, use_foreign_spelling = False):
+        assert alphabet in [Alphabet.HEPBURN, Alphabet.KUNREI, Alphabet.NIHON]
+        # If `use_foreign_spelling` is true, output will use the foreign spelling
+        #         provided in a UniDic lemma when available. For example, "カツ" will
+        #         become "cutlet" instead of "katsu".
+        import cutlet
+        self.g2p = cutlet.Cutlet(alphabet)
+        self.g2p.use_foreign_spelling = use_foreign_spelling
+        super().__init__(alphabet)
+
+    @classmethod
+    def get_lang(cls, target_lang: str) -> str:
+        """
+        Validates and returns the closest supported language code.
+
+        Args:
+            target_lang (str): The language code to validate.
+
+        Returns:
+            str: The validated language code.
+
+        Raises:
+            ValueError: If the language code is unsupported.
+        """
+        # this check is here only to throw an exception if invalid language is provided
+        return cls.match_lang(target_lang, ["ja"])
+
+    def phonemize_string(self, text: str, lang: str = "ja") -> str:
+        """
+        """
+        lang = self.get_lang(lang)
+        return self.g2p.romaji(text)
+
+class PyKakasiPhonemizer(BasePhonemizer):
+
+    def __init__(self, alphabet=Alphabet.HEPBURN):
+        assert alphabet in [Alphabet.HEPBURN, Alphabet.KANA, Alphabet.HIRA]
+        # kana, hira, hepburn
+        import pykakasi
+        self.g2p = pykakasi.kakasi()
+        super().__init__(alphabet)
+
+    @classmethod
+    def get_lang(cls, target_lang: str) -> str:
+        """
+        Validates and returns the closest supported language code.
+
+        Args:
+            target_lang (str): The language code to validate.
+
+        Returns:
+            str: The validated language code.
+
+        Raises:
+            ValueError: If the language code is unsupported.
+        """
+        # this check is here only to throw an exception if invalid language is provided
+        return cls.match_lang(target_lang, ["ja"])
+
+    def phonemize_string(self, text: str, lang: str = "ja") -> str:
+        """
+        """
+        lang = self.get_lang(lang)
+        return " ".join([
+            a[self.alphabet] for a in
+            self.g2p.convert(text)
+        ])
+
+
+if __name__ == "__main__":
+    text = "こんにちは"
+    text = "カツカレーは美味しい"
+    lang = "ja"
+
+    pho = OpenJTaklPhonemizer(kana=True)
+    pho = CutletPhonemizer()
+    pho = PyKakasiPhonemizer(Alphabet.KANA)
+    from phoonnx.phonemizers.mul import MisakiPhonemizer
+    #pho = MisakiPhonemizer()
+    print(f"\n--- Getting phonemes for '{text}' ---")
+    phonemes_cotovia = pho.phonemize(text, lang)
+    print(f"  Phonemes: {phonemes_cotovia}")

phoonnx/phonemizers/ko.py

@@ -0,0 +1,97 @@
+
+
+from phoonnx.phonemizers.base import BasePhonemizer
+from phoonnx.thirdparty.hangul2ipa import hangul2ipa
+from phoonnx.config import Alphabet
+
+
+class G2PKPhonemizer(BasePhonemizer):
+
+    def __init__(self, descriptive=True, group_vowels=True, to_syl=True,
+                 alphabet=Alphabet.IPA):
+        assert alphabet in [Alphabet.IPA, Alphabet.HANGUL]
+        from g2pk import G2p
+        self.g2p = G2p()
+        self.descriptive = descriptive
+        self.group_vowels = group_vowels
+        self.to_syl = to_syl
+        super().__init__(alphabet)
+
+    @classmethod
+    def get_lang(cls, target_lang: str) -> str:
+        """
+        Validates and returns the closest supported language code.
+
+        Args:
+            target_lang (str): The language code to validate.
+
+        Returns:
+            str: The validated language code.
+
+        Raises:
+            ValueError: If the language code is unsupported.
+        """
+        # this check is here only to throw an exception if invalid language is provided
+        return cls.match_lang(target_lang, ["ko"])
+
+    def phonemize_string(self, text: str, lang: str = "ko") -> str:
+        """
+        """
+        lang = self.get_lang(lang)
+        p = self.g2p(text, descriptive=self.descriptive,
+                     group_vowels=self.group_vowels,
+                     to_syl=self.to_syl)
+        if self.alphabet == Alphabet.IPA:
+            return hangul2ipa(p)
+        return p
+
+
+class KoG2PPhonemizer(BasePhonemizer):
+    """https://github.com/scarletcho/KoG2P"""
+    def __init__(self,   alphabet=Alphabet.IPA):
+        assert alphabet in [Alphabet.IPA, Alphabet.HANGUL]
+        from phoonnx.thirdparty.kog2p import runKoG2P
+        self.g2p = runKoG2P
+        super().__init__(alphabet)
+
+    @classmethod
+    def get_lang(cls, target_lang: str) -> str:
+        """
+        Validates and returns the closest supported language code.
+
+        Args:
+            target_lang (str): The language code to validate.
+
+        Returns:
+            str: The validated language code.
+
+        Raises:
+            ValueError: If the language code is unsupported.
+        """
+        # this check is here only to throw an exception if invalid language is provided
+        return cls.match_lang(target_lang, ["ko"])
+
+    def phonemize_string(self, text: str, lang: str = "ko") -> str:
+        """
+        """
+        lang = self.get_lang(lang)
+        p = self.g2p(text)
+        if self.alphabet == Alphabet.IPA:
+            return hangul2ipa(p)
+        return p
+
+
+if __name__ == "__main__":
+
+    pho = G2PKPhonemizer(ipa=False)
+    pho2 = KoG2PPhonemizer(ipa=False)
+    lang = "ko"
+
+    text = "터미널에서 원하는 문자열을 함께 입력해 사용할 수 있습니다."
+    print(f"\n--- Getting phonemes for '{text}' ---")
+    phonemes_cotovia = pho.phonemize(text, lang)
+    print(f"  G2PK Phonemes: {phonemes_cotovia}")
+
+    phonemes_cotovia = pho2.phonemize(text, lang)
+    print(f"  KoG2P Phonemes: {phonemes_cotovia}")
+