Semantic parsing using embedding space representations of example natural language queries
Topics: Probably in use, Search Intent, Search Query Processing
This Google patent describes a system for semantic parsing that converts natural language queries into structured data using embedding space representations. By comparing the mathematical “fingerprints” (embeddings) of a user’s query against example queries provided by developers, the system can more accurately identify the user’s intent and specific details, known as arguments. This method allows for more flexible language understanding and provides a way for developers to quickly fix errors in how the model triggers certain responses.
