LARGE LANGAUGE MODEL ARCHITECTURE TO LEVERAGE PUBLIC AND PRIVATE DATA

Smart overview of the Invention

The patent application describes a method and system for an intelligent chat powered by a large language model. This system is designed to leverage both public and private data sources to answer user queries effectively. By combining information from these diverse sources, the system aims to provide accurate and comprehensive responses to user questions, enhancing the search experience beyond traditional methods.

Functionality

Upon receiving a user query, the system executes searches against public and private data sources. Public data is accessed through search indices, while private data requires user credentials for retrieval. The large language model processes a rewritten version of the query along with context from the retrieved data, generating a natural language response that includes relevant information and links to the sources.

System Components

The system comprises memory with computer-readable instructions and processors that execute these instructions. Operations include receiving user queries, generating query rewrites, and producing responses. The model accesses both public documents using search indices and private documents using user credentials, ensuring that answers are informed by a comprehensive set of data.

Advantages

This approach addresses limitations in current search systems by integrating private data into the search process. Traditional search engines focus on public data and return a list of links for users to explore independently. In contrast, this system provides direct answers derived from both public and private data, streamlining the user's search experience and improving efficiency.

Technical Challenges

Current search systems often fail to access private data or distinguish between public and private information effectively. By incorporating both types of data, this intelligent chat system overcomes these challenges, offering users more relevant and satisfying search results. The model's ability to rewrite queries and use context enhances its capability to deliver precise answers.