I have trained Seq2Seq and LSTM models to guess the missing letters from a word.
When a word is given eg: app_e whit missing letter. the model will output the missing letter eg: app_e --> l.
The model is only trained with limited dataset. For further training I've give codes which you can use.