Leveraging Neural Machine Translation Models for Japanese to English Translation Prediction
Abstract (300 words maximum)
Neural Machine Translation (NMT) has revolutionized the field of language translation, offering unparalleled accuracy compared to traditional statistical approaches. The mid-2010's brought about a resurgence in the NMT model because of better computation power that could process these larger datasets and using CNN and RNN model alongside NMT model. This study explores the efficacy of NMT models specifically, in the context of Japanese to English translation. Using the Japanese-English Subtitle Corpus (JESC) dataset to train, we will be leveraging the learning architecture of NMT to predict our English translation. We investigate the performance of various NMT frameworks on this language pair, considering facts such as vocabulary size, model architecture, and training data size. Through experimentation and evaluation, we assess the capability of NMT models accuracy when predicting translations from Japanese to English across a diverse range of texts and linguistic nuances that are find in media and within the JESC dataset. Our findings will shed light on the strengths and limitations of NMT when performing predictive translation.
Academic department under which the project should be listed
CCSE - Computer Science
Primary Investigator (PI) Name
Md Abdullah Al Hafiz Khan
Leveraging Neural Machine Translation Models for Japanese to English Translation Prediction
Neural Machine Translation (NMT) has revolutionized the field of language translation, offering unparalleled accuracy compared to traditional statistical approaches. The mid-2010's brought about a resurgence in the NMT model because of better computation power that could process these larger datasets and using CNN and RNN model alongside NMT model. This study explores the efficacy of NMT models specifically, in the context of Japanese to English translation. Using the Japanese-English Subtitle Corpus (JESC) dataset to train, we will be leveraging the learning architecture of NMT to predict our English translation. We investigate the performance of various NMT frameworks on this language pair, considering facts such as vocabulary size, model architecture, and training data size. Through experimentation and evaluation, we assess the capability of NMT models accuracy when predicting translations from Japanese to English across a diverse range of texts and linguistic nuances that are find in media and within the JESC dataset. Our findings will shed light on the strengths and limitations of NMT when performing predictive translation.