2012 ©
             Publication
Journal Publication
Research Title Question Generation from Thai Wiki paragraphs by using Transformer 
Date of Distribution 18 March 2022 
Conference
     Title of the Conference The 2nd International Conference on Science Technology & Innovation-Maejo University  
     Organiser Faculty of Science, Maejo University, Thailand 
     Conference Place Online zoom 
     Province/State เชียงใหม่ 
     Conference Date 18 March 2022 
     To 18 March 2022 
Proceeding Paper
     Volume
     Issue
     Page 65-79 
     Editors/edition/publisher Faculty of Science, Maejo University, Thailand 
     Abstract Question Generation (QG) is one of the tasks in Natural Language Processing (NLP) that allows computers to understand the text and human language, which is important in education. The purpose of QG is to generate interrogative sentences similar to humans. However, there are very few researches related to generating Thai language questions. In this study, we proposed a method for generating a Thai wh-question (what, when, where, who, and which) with a multilingual variant of T5, a model in a Transformer architecture whose efficiency is state-of-the-art on many multilingual benchmarks covering 101 languages, including Thai.We used the iApp Wiki QA dataset for training, validation, and test. We generated questions using Thai paragraphs from the iApp Wiki QA dataset and evaluated the performance of the questions with precision and recall. Finally, we evaluated the diversity of generated questions in two methods: 1) normal input paragraphs and 2) splitting a sentence into sub-sentences. Both methods have the number of questions equal to 5788. We evaluated both methods on Self-BLEU4 and the percentage of repetition of questions. As a result, Method 1 obtained Self-BLEU4 of 19.01 and repetition of 42.03%, and Method 2 obtained Self-BLEU4 of 17.43 and repetition of 20.97% 
Author
635020064-5 Mr. NAPAT SRICHANTAPONG [Main Author]
College of Computing Master's Degree

Peer Review Status ไม่มีผู้ประเมินอิสระ 
Level of Conference นานาชาติ 
Type of Proceeding Full paper 
Type of Presentation Oral 
Part of thesis true 
Presentation awarding true 
     Award Title Announcement of the presentation awards, matchematics, Statistics & Data Science: Honorable Mention 
     Type of award รางวัลด้านวิชาการ วิชาชีพ 
     Organiser Faculty of Science, Maejo University, Thailand 
     Date of awarding 18 มีนาคม 2565 
Attach file
Citation 0