Privacy- and Utility-Preserving NLP with Anonymized Data: A case study of Pseudonymization