Text Normalization by Bi-LSTM Model with Enhanced Features to Improve Tribal English Knowledge

Education is important to all children but some children have some issues with the education for example tribal children, refugees and differently disabled children. Tribal children face some barriers in their education like language problems, they are in remote areas and economic issues also. One of the strongest reasons for their study issue is language. Language is a bridge for communication but here it is a problem for these tribal students. Because learning and teaching language is different from their mother language. For this reason, they are unable to learn anything from school. Due to this reason most, tribal students discontinue their studies, to overcome this issue they studied with their own language. Artificial Intelligence will help to eliminate this barrier from them. AI based speech recognition systems with text normalization assist them to continue their studies with the mother language of tribal children. To create a text normalizer that helps to educate tribal children with proper transcription of what they are studying. Speech recognition system by using CNN model and Bi-LSTM to create a text normalizer proposed in this paper. This convergence helps to remove grammatical errors, mistakes, time, date and frequency error from the transcription. This method achieves 92.17% accuracy range during test time and it shows better feature performances efficiently.