In ML, using RNN for an NLP project, is it necessary for DATA redundancy?

Question

Is it necessary to repeat similar template data... Like the meaning and context is the same, but the smaller details vary. If I remove these redundancies, the dataset is ver