Dynamic Document Retrieval for Enterprise AI Prompt Depth Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional search applications require users to review multiple documents to find relevant information, wasting time and computing resources, while generative AI models struggle with processing large enterprise databases efficiently.
Innovation Solution
A system that uses a depth score and relevancy scores to identify a minimal subset of documents for a generative AI model, optimizing the prompt size and reducing computational resources by incorporating only relevant content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If all documents in the database are processed and included in the prompt for the language model, then the completeness and accuracy of information retrieval is improved, but the computational resources and time required increase significantly
Solution Approach 1:
The patent extracts only the most relevant documents and content from the database based on relevancy scores and depth scores, rather than processing all documents. This selective extraction reduces computational resources while maintaining retrieval accuracy by focusing on high-value information.
Solution Approach 2:
The patent applies different processing depths to different documents based on their relevancy scores. High-relevancy documents receive deeper processing (more content extracted), while lower-relevancy documents receive shallower processing. This local quality approach optimizes resource allocation by matching processing intensity to document importance.
2Reliability
If all documents in the database are processed and included in the prompt for the language model, then the completeness and accuracy of information retrieval is improved, but the time required to process documents increases
Solution Approach 1:
The patent extracts only the most relevant documents and content from the database based on relevancy scores and depth scores, rather than processing all documents. This selective extraction reduces processing time while maintaining retrieval accuracy by focusing on high-value information.
Solution Approach 2:
The patent performs partial processing by selecting a subset of documents that are most likely to contain the answer, rather than exhaustively processing all documents. The depth score mechanism allows the system to perform just enough processing to achieve accurate results without unnecessary additional time expenditure.
3Reliability
If a larger number of documents are included in the prompt for the language model, then the accuracy of the response is improved, but the size of the prompt and number of tokens increase
Solution Approach 1:
The patent extracts only the most relevant documents and content from the database based on relevancy scores and depth scores, rather than processing all documents. This selective extraction reduces prompt size and token count while maintaining response accuracy by focusing on high-value information.
Solution Approach 2:
The patent applies different processing depths to different documents based on their relevancy scores. High-relevancy documents receive deeper processing (more content extracted), while lower-relevancy documents receive shallower processing. This local quality approach optimizes prompt size by allocating token budget according to document importance.
4Ease of operation
If traditional search applications are used, then users can find documents, but users must manually review each document which reduces efficiency and wastes computing resources
Solution Approach 1:
The patent introduces a generative AI model as an intermediary between the database and the user. This intermediary automatically processes and synthesizes information from relevant documents, eliminating the need for users to manually review each document while maintaining ease of operation through natural language interaction.
Solution Approach 2:
The system performs self-service by automatically retrieving, processing, and synthesizing information from documents without requiring user intervention. The generative AI model independently completes the information retrieval task, improving productivity while keeping the user interface simple and easy to use.
Data Source
AI summary
Systems and methods for resource-efficient retrieval of information using a generative AI model are disclosed. An input query requesting information from a set of documents is used in a prompt for a generative AI model to generate a search query to identify the documents relevant to the input query and their respective relevancy scores. The input query is used as an input another model to determine a depth score indicating a predicted number of documents needed to retrieve the information. Based on the depth score and the relevancy scores of the relevant documents, the system extracts grounding data from the identified relevant documents to generate an answer synthesis prompt for the generative AI model. The generative AI model processes the second to produce a response to the input query including the requested information.