I am building a translation pipeline for my organization using Helsinki-NLP Out Of Box translation models (no fine-tuning involved). I have a requirement where I have to retain certain PII data (like email, mobile number, and geographic data) and a few organization-based glossary words as-is in the translated text.
I understand that the quality of translation might get affected but I am ok with it.
Can someone suggest a way to achieve this with pre-trained models?