INTERNATIONAL JOURNAL OF SCIENTIFIC DEVELOPMENT AND RESEARCH International Peer Reviewed & Refereed Journals, Open Access Journal ISSN Approved Journal No: 2455-2631 | Impact factor: 8.15 | ESTD Year: 2016
open access , Peer-reviewed, and Refereed Journals, Impact factor 8.15
The purpose of this research is to study Question Answering method using XLM-RoBERTa, ByT5 and mT5 Models for closed domain network device in Lao language, the aim is to develop Question answering system to help answer question more quickly and close to human as much as possible in Lao language. Therefore, the dataset is retrieved and created from www.huawei.com by using web scraping technique, and then translated to Lao language. The dataset is in SQuAD format that consists of 921 articles (samples), where 330 articles is translated to Lao language remaining 591 articles keep in english, dataset has question 989 samples in Lao language, answer has 989 samples (393 Lao samples). The result of the research shows that training time of the XLM-RoBERTa model takes 58.3 minutes, the evaluation result by exact match is 51.51% and F1 Score is 78.38%. For the ByT5 model, training time is 120.76 minutes, evaluation result by exact match is 29.29% and F1 Score is 62.42%. The final model, the mT5 takes 82.28 minutes for training time, the exact match is 3.03% and F1 Score is 38.01%
"Study the XLM-RoBERTa, ByT5, mT5 question and answer methods for Lao language in specific areas of internet network equipment", International Journal of Science & Engineering Development Research (www.ijsdr.org), ISSN:2455-2631, Vol.8, Issue 7, page no.29 - 35, July-2023, Available :http://www.ijsdr.org/papers/IJSDR2307006.pdf
Downloads:
000338536
Publication Details:
Published Paper ID: IJSDR2307006
Registration ID:207547
Published In: Volume 8 Issue 7, July-2023
DOI (Digital Object Identifier):
Page No: 29 - 35
Publisher: IJSDR | www.ijsdr.org
ISSN Number: 2455-2631
Facebook Twitter Instagram LinkedIn