Update README.md

README.md

@@ -42,4 +42,8 @@ size_categories:
   - Removed any consecutive characters repeated more than 3 times.
   - Tokenised the text, then lemmatized it and then removed the stopwords (excluding not).
   - The `class_label` column was renamed to `label` for use with trainer API.
-- The evaluation set had ~23000 samples, while the training set had ~186k samples, i.e. a 80:10:10 (train:test:val) split.
+- The evaluation set had ~23000 samples, while the training set had ~186k samples, i.e. a 80:10:10 (train:test:val) split.
+
+### Note
+Since this dataset was preprocessed, and stopwords and punctuation marks such as "?!" were removed from it, it might be possible that in some cases
+that, the text is having incorrect labels, as the meaning changed against the original text after preprocessing.