Machine-Learned Searchable Indexes for Accurate Candidate Ranking

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing recommendation systems face inefficiencies due to separate information retrieval and machine-learned ranking stages, leading to loss of highly ranked candidates during the information retrieval stage.

Innovation Solution

A searchable index is generated based on rules derived from a machine-learned model, incorporating machine intelligence for efficient data retrieval and ranking, eliminating the need for separate stages and improving candidate selection accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If a machine-learned model is used for ranking, then recommendation accuracy is improved, but computational efficiency deteriorates

Engineering Contradiction:

Improverecommendation accuracyVSAvoidcomputational efficiency

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The patent segments the recommendation system into two distinct stages: an information retrieval stage that quickly filters candidates using efficient indexing, and a machine-learned ranking stage that applies complex models only to the filtered subset. This segmentation allows the system to maintain high accuracy where needed while preserving computational efficiency in the initial filtering phase.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary filtering through information retrieval before applying the machine-learned model. By pre-processing candidates through efficient indexing and retrieval mechanisms, the system reduces the input size for the computationally intensive ranking stage, thereby improving overall efficiency without sacrificing accuracy.

Inventive Principle:

Principle #10Preliminary action

2Productivity

If an information retrieval tool is used, then computational efficiency is improved, but recommendation accuracy deteriorates

Engineering Contradiction:

Improvecomputational efficiencyVSAvoidrecommendation accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent merges the information retrieval tool and machine-learned model into a unified two-stage recommendation system. The output of the information retrieval stage feeds directly into the ranking stage, creating a seamless pipeline that combines the efficiency of traditional retrieval with the accuracy of machine learning, thereby resolving the contradiction between speed and precision.

Inventive Principle:

Principle #5Merging (Combining)

3Device complexity

If separate information retrieval and ranking stages are used, then system modularity is improved, but candidate loss increases

Engineering Contradiction:

Improvesystem modularityVSAvoidcandidate loss

Core Design Contradiction:

Device complexityVSLoss of information

Solution Approach 1:

The patent adjusts the parameters and thresholds in the information retrieval stage to ensure that highly ranked candidates are preserved and passed to the ranking stage. By optimizing retrieval parameters such as cutoff thresholds and filtering criteria, the system maintains modularity while minimizing candidate loss, ensuring that promising candidates are not prematurely eliminated.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20250217356A1Searchable index

Publication Date: 2025.07.03 GOOGLE LLC

AI summary

Systems and techniques are disclosed for generating entries for a searchable index based on rules generated by one or more machine-learned models. The index entries can include one or more tokens correlated with an outcome and an outcome probability. A subset of tokens can be identified based on the characteristics of an event. The index may be searched for outcomes and their respective probabilities that correspond to tokens that are similar to or match the subset of tokens based on the event.