Language and justice: reconsidering harms in NLP systems and practices - Naver Labs Europe