ML Request Response Matching via Feature Vectors

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing computing systems face inefficiencies in matching request and response data due to variations in request formats, leading to incorrect mappings and requiring manual human intervention, which is prone to observational biases and unable to handle the sheer volume of data effectively.

Innovation Solution

A two-stage technique using machine learning to predict the mapping of requests to responses by applying conditions to reduce the universe of responses, computing permutations, converting data into feature vectors, and inputting them into an ML model for likelihood prediction, with an additional step for reconciliation to ensure exclusivity and continuous training based on user feedback.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If manual human intervention is used to match requests and responses, then flexibility in handling format variations is improved, but productivity and accuracy deteriorate due to observational biases and inability to handle large volumes

Engineering Contradiction:
Improveflexibility in handling format variationsVSAvoidability to handle data volume
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent replaces manual human matching (mechanical system) with an automated machine learning system that uses supervised learning algorithms to match requests and responses. The system processes data through multiple stages including candidate generation, filtering, and machine learning-based matching, eliminating human observational biases while maintaining adaptability to format variations through feature engineering and model training.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Device complexity

If traditional matching methods are used, then simplicity of the system is improved, but manufacturing precision deteriorates due to incorrect mappings

Engineering Contradiction:
Improvesystem simplicityVSAvoidmatching accuracy
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

The patent divides the matching process into distinct stages: candidate generation, filtering, feature extraction, and machine learning-based ranking. This segmentation allows each component to be optimized independently while maintaining overall system accuracy. The multi-stage approach improves matching precision without requiring complete system redesign.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces feature vectors as an intermediary representation between raw request/response data and the machine learning model. This intermediary layer transforms unstructured data into standardized features that the model can process effectively, improving matching accuracy while maintaining system modularity.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If comprehensive response filtering is applied, then measurement precision is improved, but loss of time increases due to processing large universes of responses

Engineering Contradiction:
Improvematching accuracyVSAvoidprocessing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent performs preliminary filtering of responses before applying the machine learning model. Candidate responses are pre-filtered based on basic criteria such as format matching and key field presence, reducing the universe of responses that require computationally intensive processing. This preliminary action maintains accuracy by eliminating obviously incorrect matches before detailed analysis.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent applies filtering at multiple stages with increasing stringency. Rather than applying one comprehensive filter, the system uses sequential filtering where each stage processes only the subset of data relevant to that stage's criteria, reducing overall processing time while maintaining comprehensive accuracy through cumulative filtering effects.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentUS20230297648A1Correlating request and response data using supervised learning
Publication Date: 2023.09.21 WORKDAY INC
  • US20230297648A1 patent drawing
  • US20230297648A1 patent drawing
  • US20230297648A1 patent drawing

AI summary

The example embodiments relate to matching response data to request data. in an embodiment, a method includes retrieving a request and a plurality of responses; generating packed candidates based on the request and the plurality of responses, a given packed candidate in the packed candidates including the request and a subset of the plurality of responses; generating feature vectors corresponding to the packed candidates, a given feature vector in the feature vectors including at least one aggregated feature computed based on the request and one or more corresponding responses in a respective packed candidate; inputting the feature vectors into a machine learning (ML) model, the ML model configured to output predictions corresponding to the feature vectors; selecting a feature vector from the feature vectors based on the predictions; and storing responses associated with the optimal feature vector and the request in a data storage device.