Why xlm-roberta Tokenizer split special symbol and under bar

Intermediate
1

Hello. I’m a beginner of NLP.

My question is why xlm-roberta-base Tokenizer split special symbol and under bar. For example,

tokenizer = AutoTokenizer.from_pretrained("xlm-roberta-base")
tokenizer.tokenize(',')
>>> ['▁', ',']

tokenizer.tokenize('.')
>>>  ['▁', '.']

tokenizer.tokenize('\\')
>>>['▁', '\\']

I check other special symbols but, It’s ok like ['_!']. I don’t understand why these special symbols are splited. Please why and how to solve this problem!