A Comparison of Machine Learning and Neural Network Algorithms for An Automated Thai Essay Quality Checking
Nichaphan Noiyoo, Jessada Thutkawkornpin · 2023
Checking the quality of essay writing in Thai language is still a complicated task because Thai language is very complex language in terms of punctuation, sentence structure, word repetition, spelling, commenting, and reasoning in content. Therefore, checking the quality of an essay and scoring require the reviewer's skills in reading and interpreting that make long time to review. In addition, if in reviewing process using more than one reviewer, it might affect different quality checking standards. We collected essay in Thai language which is written by student who registered paragraph writing course from The Sirindhorn Thai Language Institute of Chulalongkorn University. This work implemented LSTM model, CNN model, BERT model and WangchanBERTa model to compare the effectiveness of checking the quality of Thai essay writing. Our experimental result shows that classification analysis compiled with WangchanBERTa can achieve high accuracy up to 90%. However, CNN model compiled with classification analysis can achieve high accuracy up to 87% while compiled with regression analysis can achieve high accuracy in the range 90%. In conclusion, the system that we proposed can predict the quality of Thai essays with high accuracy. Therefore, we recommended Wangchanberta model for classification problem and CNN model for regression problem.