Abstract

Sequential data are available in many domains including weather observation data, stock price, gene sequence, and many others. Sequential data processing is very instrumental as a solution to a variety of downstream tasks in Natural Language Processing such as language translation. Translating a language to another language has become instrumental when peoples interact with other people who speak a different language. However, language translation is not an easy computation task when there is a language-resource gap. This paper presents some empirical results from the exploration of the Recurrent Neural Network (RNN), the Gated Recurrent Unit (GRU), the Long Short-term Memory (LSTM), and the Bidirectional Long Short-term Memory (Bi-LSTM) models as machine translation models from Bahasa Indonesia to the Sundanese language. These empirical results show that the Bi-LSTM achieved the highest performance (0.95 average training accuracy and 0.92 average testings BLEU score) compared to the RNN (0.93 average training accuracy and 0.91 average testings BLEU score), the GRU (0.88 average training accuracy and 0.89 average testings BLEU score), and the LSTM (0.92 average training accuracy and 0.91 average testings BLEU score) models. These results validate some previously reported studies that claim the Bi-LSTM model potentially outperforms the other recurrent neural network-based models when it is used to process sequence datasets.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call