How preprocess texts including (mentions)

Beginners
1

Is that a good idea to replace mentions like (@ABCD) in tweet posts with unk_token to preprocess a tweet to feed the BERT model?