Language Translation Using LSTM

People from all over the world are brought together by language translation to work together, share knowledge, and develop relationships. To make the process of language translation free from human error and less expensive, we came up with an automated approach. Machine learning makes the process less time-consuming and simpler. The ability of neural machine translation achieves notable advancements over archaic methods, including Rule-based and Statistical Machine-based Translation. Neural Machine-based Translation models are examined for the Hindi-English language pair based on Long Short-Term Memory Model (LSTM), which incorporates an attention-based approach. This paper discusses about the various steps involved in building an LSTM model like data pre-processing, creation of encoders and decoders, and the analysis of training and validation accuracies. We verify if LSTM performs better than a simple RNN network. Our model is found to be helpful to translate Hindi to English sentences, which can be useful in tourism, education, and entertainment industry. Future possibilities to enhance our approach have also been discussed.