Next syllables prediction system in Dzongkha using long short-term memory

Wangchuk, Karma; Riyamongkol, Panomkhawn; Waranusast, Rattapoom

doi:10.1016/j.jksuci.2021.01.001

Cited by 4 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this paper, the next Dzongkha word prediction is presented. The study conducted on Dzongkha next syllables prediction system (Wangchuk et al, 2021) reduced keystrokes for the Dzongkha typing. However, the number of keypresses required is still more.…”

Section: Discussionmentioning

confidence: 99%

“…Typewriting, exchanging texts, and documentation have been easy with the advancement of technologies. However, Dzongkha typing is difficult (Wangchuk et al, 2021). Dzongkha is the national language of Bhutan.…”

Section: Introductionmentioning

confidence: 99%

“…Their model achieved the highest F-score of 94.40%, 94.47% precision, and 94.35% recall. Furthermore, Wangchuk et al (2021) predicted the next Dzongkha syllables using LSTM. The single-layer LSTM model with 128 cells outperformed other models with the best training accuracy of 78.33%.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Dzongkha Next Words Prediction Using Bidirectional LSTM

Wangchuk¹,

Wangchuk

Namgyel

2023

bjrd

View full text Add to dashboard Cite

Dzongkha Development Commission of Bhutan (DDC) is trying to computerize Dzongkha. However, the computerization of Dzongkha poses numerous challenges. Currently, the support for Dzongkha in modern technology is limited to printing, typing, and storage. Typewriting a single Dzongkha word requires several keypresses. As a result, typing Dzongkha is tedious. In this paper, the Dzongkha word label prediction was studied. The purpose of the study was to further reduce keystrokes and make Dzongkha typing much faster. The dataset encompasses different genres curated by DDC. The dataset consisted of 10000 sentences and 4820 unique words. Next, 52150 sequences were generated using N-gram methods followed by vectorizing text using embedding techniques. Different RNN-based models were evaluated for the next Dzongkha words prediction. Two Bi-LSTM layers with 512 hidden layer neurons gave the best accuracy of 73.89% with a loss of 1.0722.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Dzongkha Next Words Prediction Using Bidirectional LSTM

Wangchuk¹,

Wangchuk

Namgyel

2023

bjrd

View full text Add to dashboard Cite

show abstract

“…To date, there has been no attempt to study English to Dzongkha translation or vice-versa. Nonetheless, next syllables prediction for Dzongkha [2], Dzongkha word segmentation [3], text to speech [4], part-of-speech (POS) tagging, and annotation of the corpus have been studied. To keep on par with the advancement of technologies and to resolve the digital divide in the country, the study on Dzongkha translation is timely.…”

Section: Introductionmentioning

confidence: 99%

Dzongkha to English translation using the neural machine translation approach

Wangchuk

Navaneethakrishnan

Jamtsho

et al. 2023

IJEECS

View full text Add to dashboard Cite

<span>In this era of technology, a communication barrier is a thing of the past. With each passing day, different types of language-based applications are being launched. There are 109 official languages Google has translated to date. However, the Dzongkha translation has not been studied. The purpose of this paper was to study Dzongkha to English translation. The parallel corpus was collected from the Dzongkha development commission of Bhutan. The dataset consisted of 53018 sentence pairs. Unique words in Dzongkha and English were 13,393 and 12,506 respectively. Different neural machine translation models were implemented. The experimental results show that the bleu score of Seq2Seq models followed a fluctuating trend. However, the bleu score of the transformer model increases gradually. It was observed that the transformer outperformed the Seq2Seq models. The highest accuracy and the lowest training loss obtained were 84.46% and 0.014858 respectively with a bleu score of 64.89.</span>

show abstract