원문정보
초록
영어
Korean grammatical particles present a persistent challenge for Vietnamese learners due to fundamental syntactic differences between the two languages. Vietnamese lacks case-marking particles, often leading to substitution errors involving eun/ neun (은/는), i/ka (이/가), e (에), and eso (에서). Traditional teaching methods offer limited success in addressing these issues. Motivated by the need for more adaptive and learner-sensitive solutions, this paper explores the fine-tuning of the LLaMA 3.2.1B language model to correct Korean particle substitution errors commonly made by beginner Vietnamese learners. A custom dataset was developed by generating simulated learner errors based on authentic sentence structures. The model was fine-tuned using Low-Rank Adaptation (LoRA) and instruction-based prompts to ensure efficiency and contextual accuracy. Evaluation on a 5,800-sentence test set demonstrated a sentence-level accuracy of 91.15%, compared to just 8.36% for the pre-trained baseline. With appropriate fine-tuning, these results endorse the capacity of large language models for providing sound grammatical corrections that are personally suited to the requirements of the learners. This technology exhibits promising potential for intelligent tutoring systems in facilitating one-to-one, real-time feedback in second language learning environments.
목차
1. INTRODUCTION
2. RELATED WORK AND BACKGROUND
2.1 Korean Particle Errors Among Vietnamese Learners
2.2 Limitations of Traditional Teaching Methods
2.3 Transitioning to LLaMA: From Traditional GEC to Learner-Focused Fine-Tuning
3. METHODOLOGY
3.1 Workflow
3.2 Method Pipeline
4. EXPERIMENTS
4.1 Preparing a Dataset to Fine-tune LLaMA
4.2 Preparing a Dataset to Fine-tune LLaMA data
4.3 Parameter-Efficient Fine-tuning via LoRA
5. RESULTS AND DISCUSSION
5.1 Training Dynamics: Analysis of Training and Validation Loss
5.2 Manual Testing and Evaluation of Model Performance Across Sentences of Varying Lengths
ACKNOWLEDGEMENT
REFERENCES