Fine-tuning Embedding Models: Sentence Transformers and Emoji Recognition
Qdrant - Vector Database & Search Engine via YouTube
Overview
Save Big on Coursera Plus. 7,000+ courses at $160 off. Limited Time Only!
Watch a 44-minute webinar demonstrating how to fine-tune embedding models to properly handle and differentiate emoji representations in text analysis. Learn an end-to-end strategy for extending a model's vocabulary while maintaining backward compatibility, eliminating the need to recompute existing embeddings. Discover how to solve the common issue where popular embedding models treat emotionally different sentences with emojis (like "I feel so 😊" and "I feel so 😔") as identical, leading to perfect but incorrect similarity scores. Master techniques for accurately capturing emotional context in social media analysis and user sentiment detection through proper emoji processing in your vector embeddings.
Syllabus
Fine-tuning an embedding model: sentence transformers getting emotional
Taught by
Qdrant - Vector Database & Search Engine