Stochastic Vector Querying for Data Record Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data record retrieval systems in documentation applications are limited by basic keyword matching, requiring exact spelling and keyword order, which hampers efficiency and the ability of entity teams to make timely and informed decisions.
Innovation Solution
A machine learning-driven system using stochastic expansion data querying, where data records and query statements are represented as vectors, with test vectors generated around the query vector to iteratively refine the search, ultimately providing results to a large language model for accurate retrieval.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If basic keyword matching is used for data record retrieval, then the system is simple and easy to implement, but the retrieval accuracy is poor and requires exact spelling and keyword order
Solution Approach 1:
The patent transforms the retrieval process from exact keyword matching to vector-based similarity scoring. Data records and queries are converted to vectors, and a similarity function calculates relevance scores, allowing flexible matching that doesn't require exact spelling or keyword order while maintaining system manageability
Solution Approach 2:
The patent replaces the mechanical keyword-matching system with a machine learning-based vector similarity system. This substitution enables the system to understand semantic relationships and retrieve relevant records without requiring precise keyword correspondence, significantly improving retrieval accuracy
2Measurement precision
If manual review of all data records is performed, then comprehensive analysis is possible, but the process is time-consuming and resource-intensive
Solution Approach 1:
The patent implements an automated retrieval system that performs comprehensive data analysis without human intervention. The machine learning model automatically processes queries, calculates similarity scores for all records, and returns relevant results, eliminating the need for manual review while maintaining comprehensive analysis capabilities
Solution Approach 2:
The system uses similarity scoring as feedback to automatically rank and filter data records. By calculating similarity scores for all records and using this feedback to prioritize results, the system achieves comprehensive analysis efficiently without requiring manual review of every record
3Loss of information
If exact keyword matching is required, then the search process is simple, but the ability to extract valuable insights is hindered
Solution Approach 1:
The patent changes the matching parameter from exact keyword correspondence to vector similarity. This allows the system to capture semantic meaning and extract valuable insights from records that may not contain exact keyword matches, while the interface remains simple for users
Data Source
AI summary
Systems, computer program products, and methods are described herein for machine learning-driven data record retrieval via stochastic expansion data querying. Data records and a query statement are received. Using a weighting model, a data vector for each of the data records and a query vector for the query statement are generated. A first similarity score for each data vector is determined. Rankings for the data records are generated. Based on the rankings, a grouping of data records is selected. Test vectors centered from the query vector are then generated stochastically. A second similarity score for each of the test vectors is determined. A subsequent query vector is determined based on a highest second similarity score. If metrics do not satisfy a predetermined stopping criteria, subsequent test vectors are generated stochastically. Ranked data records are then provided to a large language model.