|
--- |
|
license: apache-2.0 |
|
task_categories: |
|
- question-answering |
|
language: |
|
- th |
|
tags: |
|
- exam |
|
pretty_name: OpenThaiGPT Evaluation Dataset |
|
size_categories: |
|
- n<1K |
|
--- |
|
|
|
# OpenThaiGPT Evaluation Dataset |
|
|
|
[](https://opensource.org/licenses/Apache-2.0) |
|
|
|
## Overview |
|
|
|
The **OpenThaiGPT Evaluation Dataset** is a collection of curated datasets designed for evaluating Thai-language question-answering models. The dataset is focused on benchmark exams, offering a comprehensive set of evaluation resources for various educational and professional domains in Thailand. |
|
|
|
## Available Benchmark Datasets |
|
|
|
This dataset includes multiple benchmark exams, covering both general academic and specialized professional subjects in Thailand: |
|
|
|
- **A-Level**: Standardized test for high school students in Thailand. |
|
- **TGAT**: Thai General Aptitude Test. |
|
- **TPAT1**: Thai Professional Aptitude Test, a critical evaluation for professional pathways. |
|
- **Thai Investment Consultant**: Designed to evaluate knowledge related to investment consulting in Thailand. |
|
- **Facebook Belebele Thai**: A multilingual dataset for natural language understanding in the Thai context. |
|
- **xcopa_th_200**: Cross-lingual transfer learning dataset for Thai. |
|
- **xnli2.0_th_200**: Cross-lingual natural language inference dataset for Thai. |
|
- **Thai ONET M3**: Standardized test for grade 9 students in Thailand. |
|
- **Thai ONET M6**: Standardized test for grade 12 students in Thailand. |
|
|
|
### ONET M3 (2021) |
|
- **Subject**: Thai, Social Studies, Mathematics, Science, English |
|
- **Source**: OCR from Public Exam, National Institute of Educational Testing Service (NIETS) |
|
- **Dataset Repository**: OCR and cleaned by iApp Team, donated to OpenThaiGPT. |
|
- [ONET M3 2021 Exam Papers](https://www.niets.or.th/th/content/uploads/editor/files/O-NET/exam_m3_2564/) |
|
|
|
### ONET M6 (2021-2025) |
|
- **Subject**: Thai, Mathematics, Social Studies, Science, English |
|
- **Source**: OCR from Public Exam, National Institute of Educational Testing Service (NIETS) |
|
- **Dataset Repository on Huggingface**: [Thai ONET Exam Dataset](https://huggingface.co/datasets/openthaigpt/thai-onet-exam) |
|
- [ONET M6 2021 Exam Papers](https://www.niets.or.th/th/content/uploads/editor/files/O-NET/exam_m6_2564/) |
|
|
|
### A-Level (2021-2022) |
|
- **Subjects**: Applied Mathematics, Biology, English, Social Studies |
|
- **Source**: Closed-source PDFs purchased from private tutors |
|
|
|
### TGAT (2021) |
|
- **Subjects**: Critical Thinking |
|
- **Source**: Closed-source PDFs purchased from private tutors |
|
|
|
### TPAT1 (2020-2021) |
|
- **Subjects**: Medical Ethics |
|
- **Source**: Closed-source PDFs purchased from private tutors |
|
|
|
### Thai Investment Consultant Licensing Exams |
|
- **Subjects**: IC Plain, IC Complex (P2, P3) |
|
- **Source**: OCR from Public Exam, The Stock Exchange of Thailand (SET) |
|
- **Dataset Repository on Huggingface**: [Thai Investment Consultant Licensing Exams](https://huggingface.co/datasets/openthaigpt/thai-investment-consultant-licensing-exams) |
|
|
|
### Facebook Belebele Thai (200 Rows) |
|
- **Source**: Open dataset from Huggingface |
|
- **Dataset Repository on Huggingface**: [Facebook Belebele](https://huggingface.co/datasets/facebook/belebele) |
|
- **Note**: The first 200 rows are selected for evaluation speed and cost efficiency. Cleaned by OpenThaiGPT. |
|
|
|
### XCOPA TH (200 Rows) |
|
- **Source**: Open dataset from Huggingface |
|
- **Dataset Repository on Huggingface**: [XCOPA Thai](https://huggingface.co/datasets/xcopa) |
|
- **Note**: The first 200 rows are selected for evaluation speed and cost efficiency. Cleaned by OpenThaiGPT. |
|
|
|
### XNLI 2.0 TH (200 Rows) |
|
- **Source**: Open dataset from Huggingface |
|
- **Dataset Repository on Huggingface**: [XNLI Thai](https://huggingface.co/datasets/xnli) |
|
- **Note**: The first 200 rows are selected for evaluation speed and cost efficiency. Cleaned by OpenThaiGPT. |
|
|
|
## Dataset Details |
|
|
|
Each dataset in this collection is designed to assess different facets of Thai language understanding and knowledge in various educational and professional sectors. The dataset is invaluable for training, fine-tuning, and evaluating question-answering systems and other language models in Thai. |
|
|
|
For more detailed information about each exam, please refer to the [exam details sheet](https://docs.google.com/spreadsheets/d/1ZtP5Jkx0IvCWNPQhMKitZszGnLKqvEDEf0OKdmQiXjA/edit#gid=1181424412). |
|
|
|
### Key Features |
|
- **Question Format**: All questions are provided in text format. |
|
- **Multiple Choice Questions**: The dataset includes multiple choice questions, offering a range of possible answers for each query. |
|
- **Solutions**: Correct answers to all questions are provided. |
|
- **Thai Human Verification**: Each item in the dataset has been verified by a Thai-speaking individual who is not a domain expert, ensuring the clarity and accessibility of the content. |
|
- **Extra Annotations**: |
|
- `isAnswerable`: Indicates whether the question can be answered with the provided text alone, without the need for additional information such as visual aids. |
|
- `isMultipleChoice`: Identifies if the question is a multiple choice question. |
|
- `isSingleChoiceSolution`: Specifies if there is only one correct answer among the provided choices. |
|
|
|
## Dataset Statistics |
|
|
|
- **Task Categories**: Question Answering |
|
- **Language**: Thai |
|
- **Dataset Tags**: Exam, Benchmark |
|
- **Size**: Less than 1K examples |
|
|
|
## License |
|
|
|
This dataset is distributed under the [Apache-2.0 License](https://opensource.org/licenses/Apache-2.0). |
|
|
|
## Maintainer |
|
Dr. Kobkrit Viriyayudhakorn |
|
Email: [email protected] |
|
|
|
--- |
|
|
|
This dataset is continuously updated and maintained. Contributions and feedback are welcome to improve the coverage and quality of this resource. |
