A Simple but Effective Way to Improve the Performance of RNN-Based Encoder in Neural Machine Translation Task

Liu, Yan; Zhang, Dongwen; Du, Lei; Gu, Zhaoquan; Qiu, Jing; Tan, Qingji

doi:10.1109/dsc.2019.00069

Cited by 5 publications

(3 citation statements)

References 16 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…RNN is a method that is often used in sequential data processing such as text processing and others [28]. An RNN represents a type of neural machine translation (NMT) system with three layers: an initial layer that assigns each word to a vector, like a word embedding or a one-hot word index; a looping hidden layer that continually computes and modifies the hidden state as it processes each word; and a final layer that predicts the likelihood of upcoming words while retaining the current hidden state [29].…”

Section: Training Corpus Using Rnn-grumentioning

confidence: 99%

Recurrent Neural Network-Gated Recurrent Unit for Indonesia-Sentani Papua Machine Translation

Achmad,

Tokoro,

Haurissa

et al. 2023

J. Inf. Syst. Informatics

View full text Add to dashboard Cite

The Papuan Sentani language is spoken in the city of Jayapura, Papua. The law states the need to preserve regional languages. One of them is by building an Indonesian-Sentani Papua translation machine. The problem is how to build a translation machine and what model to choose in doing so. The model chosen is Recurrent Neural Network – Gated Recurrent Units (RNN-GRU) which has been widely used to build regional languages in Indonesia. The method used is an experiment starting from creating a parallel corpus, followed by corpus training using the RNN-GRU model, and the final step is conducting an evaluation using Bilingual Evaluation Understudy (BLEU) to find out the score. The parallel corpus used contains 281 sentences, each sentence has an average length of 8 words. The training time required is 3 hours without using a GPU. The result of this research was that a fairly good BLEU score was obtained, namely 35.3, which means that the RNN-GRU model and parallel corpus produced sufficient translation quality and could still be improved.

show abstract

Section: Training Corpus Using Rnn-grumentioning

confidence: 99%

Recurrent Neural Network-Gated Recurrent Unit for Indonesia-Sentani Papua Machine Translation

Achmad,

Tokoro,

Haurissa

et al. 2023

J. Inf. Syst. Informatics

View full text Add to dashboard Cite

show abstract

“…Lapisan Encoder-Decoder terdiri dari sebuah jaringan proses pembelajaran dengan metode Recurrent Neural Network (RNN) [11]. RNN merupakan sebuah metode yang sering digunakan dalam pengolahan data sekuensial seperti pengolahan teks dan lainnya [12]. RNN memiliki salah satu pengembangan variasi yaitu Long Short-Term Memory (LSTM) yang memiliki struktur relatif kompleks dengan penggunaan sebuah sel memori untuk menghubungkan setiap data masukan pertama dan seterusnya.…”

Section: Pendahuluanunclassified

“…Matriks BLEU dirancang untuk mengukur seberapa dekat keluaran yang dihasilkan dengan melakukan pencocokan panjang frasa variabel keluaran dari mesin penterjemah dengan referensi terjemahan. Matriks dasar memerlukan sebuah kalkulasi brevity penalty dengan perhitungan pada Persamaan (12) dan Persamaan (13).…”

Section: Bleu Scoreunclassified

Mesin Penterjemah Bahasa Indonesia-Bahasa Sunda Menggunakan Recurrent Neural Networks

Fauziyah¹,

Ilyas²,

Kasyidi³

2022

JTI

View full text Add to dashboard Cite

Translator is a process where one language is changed into another language. Translator in the last research was carried out using a Phrase-based Statistical Machine Translation (PSMT) approach. This research builds an Indonesian to Sundanese translator. The stages used start from pre-processing using text preprocessing and word embedding Word2Vec and the approach used is Neural Machine Translation (NMT) with Encoder-Decoder architecture in which there is a Recurrent Neural Network (RNN). Tests in the study resulted in the optimal value by the GRU of 99.17%. Models using Attention got 99.94%. The use of optimization model got optimal results by Adam 99.35% and BLEU Score results with optimal bleu 92.63% and brievity penalty 0.929. The results of the machine translator produce training predictions from Indonesian to Sundanese if the input sentences are in accordance with the corpus and the translation results are not suitable when the input sentences are different from the corpus.

show abstract

Stock Price Analysis and Prediction Using Seq2Seq LSTM

Dash

Singh

Jain

et al. 2023

Lecture Notes in Networks and Systems

View full text Add to dashboard Cite

A Simple but Effective Way to Improve the Performance of RNN-Based Encoder in Neural Machine Translation Task

Cited by 5 publications

References 16 publications

Recurrent Neural Network-Gated Recurrent Unit for Indonesia-Sentani Papua Machine Translation

Recurrent Neural Network-Gated Recurrent Unit for Indonesia-Sentani Papua Machine Translation

Mesin Penterjemah Bahasa Indonesia-Bahasa Sunda Menggunakan Recurrent Neural Networks

Stock Price Analysis and Prediction Using Seq2Seq LSTM

Contact Info

Product

Resources

About