Thaigovai : a framework to transform into informal text to government writing style | |
| Author | Arnajak Tungchoksongchai |
| Call Number | AIT Thesis no.DSAI-23-07 |
| Subject(s) | Natural language processing (Computer science) Artificial intelligence Government report writing--Thailand--Data processing |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Data Science and Artificial Intelligence |
| Publisher | Asian Institute of Technology |
| Abstract | The background of this research is rooted in the observation that most Thai people, includ ing government officers, face challenges when writing government-style documents due to the unique style, rule-following, and use of non-regular words. Developing an application to address this issue could significantly save time and improve efficiency for individu als. However, the problem arises from the absence of a dataset for fine-tuning language models, leading to risks of hallucination in sample generation and requiring significant human effort for manual dataset creation. Additionally, implementing zero-shot learning with large language models incurs substantial costs, and existing evaluation metrics are not fully suitable for the task. To tackle these challenges, the research proposes a framework that distills knowledge from language models, filters data generation, and fine-tunes the model, all while incorporating human evaluation. The key findings reveal that datasets generated from GPT3.5 are of acceptable quality, but filtering them before fine-tuning can enhance model performance. The frameworkâs outcomes are comparable to those of non-specialists. And to the best of our knowledge, it could represent the first application of paraphrasing casual text into government style, which makes it a beneficial initiator for this innovation. |
| Year | 2023 |
| Type | Thesis |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Data Science and Artificial Intelligence (DSAI) |
| Chairperson(s) | Chaklam Silpasuwanchai; |
| Examination Committee(s) | Dailey, Matthew N.;Mongkol Ekpanyapong; |
| Scholarship Donor(s) | Royal Thai Government Fellowship; |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2023 |