QUESTION GENERATION FROM MATHEMATICS CONTEXTUAL TEST USING GOOGLE T5
This study explores the application of natural language processing (NLP) to automate the generation of mathematics word problems. Specifically, it focuses on fine-tuning the pre-trained language model, idT5, to produce mathematics word problems from contextual text. The motivation behind this rese...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/84205 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | This study explores the application of natural language processing (NLP) to automate the generation of mathematics word problems. Specifically, it focuses on fine-tuning the pre-trained language model, idT5, to produce mathematics word problems from contextual text. The motivation behind this research stems from the shift towards competency-based assessments, such as Indonesia's National Assessment, which emphasize real-world problem-solving skills. By leveraging the capabilities of the idT5 model, this study aims to streamline the process of creating assessment items, particularly those involving mathematical literacy. The study involves translating and adapting the GSM8K mathematics word problem dataset into Indonesian, followed by fine-tuning the idT5 model on this adapted dataset. The model's performance is evaluated using the ROUGE score to measure its ability to generate high-quality, contextually relevant mathematics problems. Experimental results indicate a satisfactory ROUGE score, suggesting that the fine-tuned model can effectively perform question generation. Human evaluation further confirms that the generated questions are mathematically contextual. While the model shows promising results, there is still room for improvement, such as increasing the dataset size and exploring alternative evaluation metrics.
|
|---|