Tokenization is Sensitive to Language Variation

Open in new window