Latest advances in deep studying and computerized speech recognition have boosted the accuracy of end-to-end speech recognition to a brand new degree. Nevertheless, recognition of private content material resembling contact names stays a problem. On this work, we current a personalization resolution for an end-to-end system based mostly on connectionist temporal classification. Our resolution makes use of class-based language mannequin, during which a normal language mannequin offers modeling of the context for named entity courses, and private named entities are compiled in a separate finite state transducer. We additional introduce a phoneme-to-wordpeice mannequin to map uncommon named entities to extra frequent homophonic wordpieces, and in addition wordpiece prior normalization to bias for uncommon wordpieces, main to a different 48.9% relative enchancment in private named entity accuracy on prime of an already personalised baseline. This work permits our techniques to match extremely aggressive personalised hybrid techniques on private named entity recognition.