AI & Tech··1 sources verified
Airbnb Unveils CASTLE to Train Natural Language Search Models on Synthetic Data
Summarised by Relevant News AI · Read time: 3 min
Airbnb researchers introduced CASTLE, a framework that tackles the cold-start problem in natural language search. It uses de-identified booking sessions and large language models to jointly generate synthetic search queries and relevance labels. This lets teams train search models before they have enough real user search traffic.
Why it matters: It offers a practical method for ML teams to bootstrap LLM-based search relevance models using synthetic data when real query logs are scarce.
All sources