Algorithm Intern (AI text data annotation, fine-tuning)
Text
Fine-tuningBuilt and fine-tuned a BERT-based model for sentiment analysis and topic extraction using large-scale financial news text data. Responsible for automating data collection, cleaning, and preparing labeled datasets for multi-task learning, including downstream sentiment recognition and topic classification tasks. Implemented end-to-end data pipeline ensuring high-quality labeled data and improved algorithm output accuracy. • Curated and annotated over 100,000 news articles for emotion and topic labels. • Designed detailed annotation guidelines and validated consistency across annotators. • Performed quality assurance by reviewing and correcting model-generated label results. • Used PyTorch as the primary software for data management and fine-tuning tasks.