[Whisper] Allow basic text normalization (#26149)

* [Whisper] Allow basic text normalization

* up

* style copies

src/transformers/models/whisper/tokenization_whisper.py

@@ -23,7 +23,7 @@
 
 from ...tokenization_utils import AddedToken, PreTrainedTokenizer
 from ...utils import logging
-from .english_normalizer import EnglishTextNormalizer
+from .english_normalizer import BasicTextNormalizer, EnglishTextNormalizer
 
 
 VOCAB_FILES_NAMES = {
@@ -510,6 +510,15 @@ def _normalize(self, text):
         normalizer = EnglishTextNormalizer(self.english_spelling_normalizer)
         return normalizer(text)
 
+    @staticmethod
+    def _basic_normalize(text, remove_diacritics=False):
+        """
+        Normalize a given string using the `BasicTextNormalizer` class, which preforms commons transformation on
+        multilingual text.
+        """
+        normalizer = BasicTextNormalizer(remove_diacritics=remove_diacritics)
+        return normalizer(text)
+
     def _decode_with_timestamps(self, token_ids, skip_special_tokens=False, time_precision=0.02) -> str:
         """
         Timestamp tokens are above the special tokens' id range and are ignored by `decode()`. This method decodes
@@ -617,6 +626,9 @@ def decode(
         output_offsets: bool = False,
         time_precision=0.02,
         decode_with_timestamps: bool = False,
+        normalize: bool = False,
+        basic_normalize: bool = False,
+        remove_diacritics: bool = False,
         **kwargs,
     ) -> str:
         """
@@ -633,15 +645,24 @@ def decode(
             clean_up_tokenization_spaces (`bool`, *optional*):
                 Whether or not to clean up the tokenization spaces. If `None`, will default to
                 `self.clean_up_tokenization_spaces` (available in the `tokenizer_config`).
-            kwargs (additional keyword arguments, *optional*):
-                Will be passed to the underlying model specific decode method.
             output_offsets (`bool`, *optional*, defaults to `False`):
                 Whether or not to output the offsets of the tokens. This should only be set if the model predicted
                 timestamps.
             time_precision (`float`, `optional`, defaults to 0.02):
                 The time ratio to convert from token to time.
             decode_with_timestamps (`bool`, *optional*, defaults to `False`):
                 Whether or not to decode with timestamps included in the raw text.
+            normalize (`bool`, *optional*, defaults to `False`):
+                Whether or not to apply the English text normalizer to the decoded text. Only applicable when the
+                target text is in English. Otherwise, the basic text normalizer should be applied.
+            basic_normalize (`bool`, *optional*, defaults to `False`):
+                Whether or not to apply the Basic text normalizer to the decoded text. Applicable to multilingual
+                target text.
+            remove_diacritics (`bool`, *optional*, defaults to `False`):
+                Whether or not to remove diacritics when applying the Basic text normalizer. Removing diacritics may
+                destroy information in the decoded text, hence it should be used with caution.
+            kwargs (additional keyword arguments, *optional*):
+                Will be passed to the underlying model specific decode method.
         Returns:
             `str`: The decoded sentence.
         """
@@ -654,7 +675,9 @@ def decode(
             filtered_ids,
             skip_special_tokens=skip_special_tokens,
             clean_up_tokenization_spaces=clean_up_tokenization_spaces,
-            decode_with_timestamps=decode_with_timestamps,
+            normalize=normalize,
+            basic_normalize=basic_normalize,
+            remove_diacritics=remove_diacritics,
             **kwargs,
         )
         if decode_with_timestamps:
@@ -676,7 +699,8 @@ def _decode(
         token_ids: Union[int, List[int]],
         skip_special_tokens: bool = False,
         normalize: bool = False,
-        decode_with_timestamps: bool = False,
+        basic_normalize: bool = False,
+        remove_diacritics: bool = False,
         **kwargs,
     ) -> str:
         self._decode_use_source_tokenizer = kwargs.pop("use_source_tokenizer", False)
@@ -705,6 +729,9 @@ def _decode(
         if normalize:
             clean_text = self._normalize(text)
             return clean_text
+        elif basic_normalize:
+            clean_text = self._basic_normalize(text, remove_diacritics=remove_diacritics)
+            return clean_text
         else:
             return text
 

src/transformers/models/whisper/tokenization_whisper_fast.py

@@ -25,7 +25,7 @@
 from ...tokenization_utils_base import BatchEncoding
 from ...tokenization_utils_fast import PreTrainedTokenizerFast
 from ...utils import logging
-from .english_normalizer import EnglishTextNormalizer
+from .english_normalizer import BasicTextNormalizer, EnglishTextNormalizer
 from .tokenization_whisper import LANGUAGES, TASK_IDS, TO_LANGUAGE_CODE, WhisperTokenizer, _decode_asr
 
 
@@ -331,6 +331,9 @@ def decode(
         output_offsets: bool = False,
         time_precision=0.02,
         decode_with_timestamps: bool = False,
+        normalize: bool = False,
+        basic_normalize: bool = False,
+        remove_diacritics: bool = False,
         **kwargs,
     ) -> str:
         """
@@ -347,15 +350,24 @@ def decode(
             clean_up_tokenization_spaces (`bool`, *optional*):
                 Whether or not to clean up the tokenization spaces. If `None`, will default to
                 `self.clean_up_tokenization_spaces` (available in the `tokenizer_config`).
-            kwargs (additional keyword arguments, *optional*):
-                Will be passed to the underlying model specific decode method.
             output_offsets (`bool`, *optional*, defaults to `False`):
                 Whether or not to output the offsets of the tokens. This should only be set if the model predicted
                 timestamps.
             time_precision (`float`, `optional`, defaults to 0.02):
                 The time ratio to convert from token to time.
             decode_with_timestamps (`bool`, *optional*, defaults to `False`):
                 Whether or not to decode with timestamps included in the raw text.
+            normalize (`bool`, *optional*, defaults to `False`):
+                Whether or not to apply the English text normalizer to the decoded text. Only applicable when the
+                target text is in English. Otherwise, the basic text normalizer should be applied.
+            basic_normalize (`bool`, *optional*, defaults to `False`):
+                Whether or not to apply the Basic text normalizer to the decoded text. Applicable to multilingual
+                target text.
+            remove_diacritics (`bool`, *optional*, defaults to `False`):
+                Whether or not to remove diacritics when applying the Basic text normalizer. Removing diacritics may
+                destroy information in the decoded text, hence it should be used with caution.
+            kwargs (additional keyword arguments, *optional*):
+                Will be passed to the underlying model specific decode method.
         Returns:
             `str`: The decoded sentence.
         """
@@ -368,7 +380,9 @@ def decode(
             filtered_ids,
             skip_special_tokens=skip_special_tokens,
             clean_up_tokenization_spaces=clean_up_tokenization_spaces,
-            decode_with_timestamps=decode_with_timestamps,
+            normalize=normalize,
+            basic_normalize=basic_normalize,
+            remove_diacritics=remove_diacritics,
             **kwargs,
         )
         if decode_with_timestamps:
@@ -385,12 +399,17 @@ def decode(
             return {"text": text, "offsets": offsets}
         return text
 
-    def _decode(self, *args, normalize: bool = False, **kwargs) -> str:
+    def _decode(
+        self, *args, normalize: bool = False, basic_normalize: bool = False, remove_diacritics: bool = False, **kwargs
+    ) -> str:
         text = super()._decode(*args, **kwargs)
 
         if normalize:
             clean_text = self._normalize(text)
             return clean_text
+        elif basic_normalize:
+            clean_text = self._basic_normalize(text, remove_diacritics=remove_diacritics)
+            return clean_text
         else:
             return text
 
@@ -403,6 +422,16 @@ def _normalize(self, text):
         normalizer = EnglishTextNormalizer(self.english_spelling_normalizer)
         return normalizer(text)
 
+    @staticmethod
+    # Copied from transformers.models.whisper.tokenization_whisper.WhisperTokenizer._basic_normalize
+    def _basic_normalize(text, remove_diacritics=False):
+        """
+        Normalize a given string using the `BasicTextNormalizer` class, which preforms commons transformation on
+        multilingual text.
+        """
+        normalizer = BasicTextNormalizer(remove_diacritics=remove_diacritics)
+        return normalizer(text)
+
     def save_vocabulary(self, save_directory: str, filename_prefix: Optional[str] = None) -> Tuple[str]:
         files = self._tokenizer.model.save(save_directory, name=filename_prefix)
 

tests/models/whisper/test_tokenization_whisper.py

@@ -273,6 +273,40 @@ def test_combine_tokens_into_words(self):
         self.assertEqual(expected_tokens, output_rust[1])
         self.assertEqual(expected_indices, output_rust[2])
 
+    def test_basic_normalizer(self):
+        tokenizer = self.get_tokenizer()
+        rust_tokenizer = self.get_rust_tokenizer()
+
+        input_str = "Hola güey!"
+        expected_output_normalize = "hola güey "
+        expected_output_diacritics = "hola guey "
+
+        # tokenizer tests
+        encoded_input = tokenizer(input_str).input_ids
+        decoded_output = tokenizer.decode(encoded_input, skip_special_tokens=True, basic_normalize=False)
+        self.assertEqual(decoded_output, input_str)
+
+        decoded_output_normalize = tokenizer.decode(encoded_input, skip_special_tokens=True, basic_normalize=True)
+        self.assertEqual(decoded_output_normalize, expected_output_normalize)
+
+        decoded_output_diacritics = tokenizer.decode(
+            encoded_input, skip_special_tokens=True, basic_normalize=True, remove_diacritics=True
+        )
+        self.assertEqual(decoded_output_diacritics, expected_output_diacritics)
+
+        # fast tokenizer tests
+        encoded_input = rust_tokenizer(input_str).input_ids
+        decoded_output = rust_tokenizer.decode(encoded_input, skip_special_tokens=True, basic_normalize=False)
+        self.assertEqual(decoded_output, input_str)
+
+        decoded_output_normalize = rust_tokenizer.decode(encoded_input, skip_special_tokens=True, basic_normalize=True)
+        self.assertEqual(decoded_output_normalize, expected_output_normalize)
+
+        decoded_output_diacritics = rust_tokenizer.decode(
+            encoded_input, skip_special_tokens=True, basic_normalize=True, remove_diacritics=True
+        )
+        self.assertEqual(decoded_output_diacritics, expected_output_diacritics)
+
 
 class SpeechToTextTokenizerMultilinguialTest(unittest.TestCase):
     checkpoint_name = "openai/whisper-small.en"