BERT Traditional/Simplified Chinese Classifier
BERT-based classifier for Traditional Chinese variants (Taiwan vs Mainland). Achieves 87.71% accuracy with long-text support, Focal Loss, and MC Dropout voting.
Showcasing my past projects spanning NLP, Deep Learning, and AI applications.
BERT-based classifier for Traditional Chinese variants (Taiwan vs Mainland). Achieves 87.71% accuracy with long-text support, Focal Loss, and MC Dropout voting.
Established the 'Formosa Vision Dialogue Corpus' using images from National Cultural Memory Bank 2.0 to enhance VLM's understanding of Taiwan's landmarks, history, and culture.
A resume matching system balancing performance and fairness, combining LoRA fine-tuning, adversarial debiasing, and XAI for transparent decision-making.
A Chinese SMS classification system based on multi-stage clustering and contrastive learning, using SimCSE-style methods to fine-tune Sentence Encoder.
High-performance CLI tool leveraging local Ollama models to generate and augment training datasets for LLM fine-tuning with structured JSON output and concurrent batch processing.
Smart travel planning system combining Pandas AI and GPT-4o RAG retrieval, supporting multimodal input and customizable requirements.