Dynamic Document Retrieval Using Depth Scoring for Enterprise AI
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 generative AI model to analyze enterprise databases, identifying a depth score and relevancy scores to select a minimal subset of documents for processing, optimizing the prompt size and reducing computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If all documents in the database are processed and included in the AI prompt, then the completeness of information is improved, but the computational resources and time required increase significantly
Solution Approach 1:
The patent segments the large database into individual documents and further segments the processing into multiple stages: initial search query generation, relevance scoring, depth score calculation, and selective extraction. This segmentation allows the system to process only necessary portions of the database rather than all documents simultaneously, reducing computational resource consumption while maintaining information completeness.
Solution Approach 2:
The patent extracts only the most relevant documents and specific content portions from the larger database based on calculated relevance scores and depth scores. By taking out only the essential information needed to answer the query rather than processing the entire database, the system significantly reduces computational resources and prompt size while preserving the completeness of relevant information.
2Reliability
If multiple documents are reviewed to ensure accurate information retrieval, then the accuracy of results is improved, but the time required to find relevant information increases
Solution Approach 1:
The patent performs preliminary actions by generating an initial search query, retrieving candidate documents, and calculating both relevance scores and depth scores before the final information retrieval. This preliminary filtering and scoring process identifies and prioritizes the most relevant documents in advance, allowing the system to focus subsequent processing on a smaller, pre-selected set of high-value documents, thereby maintaining accuracy while reducing time.
3Manufacturing precision
If a larger number of tokens are used in the prompt to include more document content, then the quality of the language model response is improved, but the processing time and resource consumption increase
Solution Approach 1:
The patent applies local quality by differentiating the treatment of different documents based on their relevance scores and depth scores. High-relevance documents with appropriate depth scores receive full or partial inclusion in the prompt, while lower-relevance documents are excluded or summarized. This selective approach ensures that the prompt contains sufficient high-quality content from the most relevant sources without being bloated with unnecessary tokens from less relevant documents, thereby maintaining response quality while improving processing speed.
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.