Named entity recognition (NER) is a crucial Natural language processing (NLP) task used in applications like voice assistants, search engines, customer support, etc. A lack of entities relevant to the use case makes the available datasets insufficient for training. Data augmentation is a method in which synthetic data is fabricated from existing data to enhance the existing dataset. The existing data augmentation methods do not consider the grammatical and logical correctness of the fabricated sentences, resulting in a decrease in the performance of transformer-based NER models. This paper proposes a novel data augmentation method with a sanity-checker that checks the correctness of the augmented sentences and produces augmented data that improves the performance of transformer-based NER models. When the proposed augmentation algorithm was tested with the CoNLL-2003 dataset, a significant increase in the F1 score of BERT based NER from 94.73 to 95.37% and RoBERTa based NER from 94.13 to 95.14% was observed.