Create normalize.py

normalize.py

@@ -0,0 +1,44 @@
+import re
+import emoji
+
+def normalize(sentence: str) -> str:
+    """
+    This function should be used before tokenizing the input string.
+
+    Normalizes the input string in the following ways:
+    -> Converts from ş to ș, ţ to ț, etc.
+    -> Converts @mention to USER, #hashtag to HASHTAG, http... and www... to HTTPURL
+    -> Converts emoticons to :emoji_with_long_name:
+    -> Replaces :emoji_with_long_name: with emoji_with_long_name and replaces _, : and - with empty string
+    -> Removes multiple whitespaces with a single whitespace
+    Args:
+        `sentence` (str): The input string.
+    Returns:
+        str: The normalized string.
+    """
+
+    # Make sure it's a string
+    sentence = str(sentence)
+
+    # Convert from ş to ș, ţ to ț, etc.
+    sentence = re.sub(r'ş', 'ș', sentence)
+    sentence = re.sub(r'Ş', 'Ș', sentence)
+    sentence = re.sub(r'ţ', 'ț', sentence)
+    sentence = re.sub(r'Ţ', 'Ț', sentence)
+
+    # Convert @mentions to USER, #hashtags to HASHTAG, http... and www... to HTTPURL
+    sentence = re.sub(r'@\S+', 'USER', sentence)
+    sentence = re.sub(r'#\S+', 'HASHTAG', sentence)
+    sentence = re.sub(r'http\S+', 'HTTPURL', sentence)
+    sentence = re.sub(r'www\S+', 'HTTPURL', sentence)
+
+    # Convert emoticons to :emoji_with_long_name:
+    sentence = emoji.demojize(sentence, delimiters=(' :', ': '))
+
+    # Replace :emoji_with_long_name: with emojiwithlongname
+    sentence = re.sub(r':\S+:', lambda x: x.group(0).replace('_', '').replace(':', '').replace('-', ''), sentence)
+
+    # Remove multiple whitespaces with a single whitespace
+    sentence = re.sub(r'\s+', ' ', sentence)
+
+    return sentence