Machine Learning Anchor Detection for Executable Action Suggestions

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing digital assistants lack the ability to intelligently suggest executable actions based on user behavior patterns detected through electronic device interactions, leading to inefficiencies in user experience.

Innovation Solution

A predictive model is trained to detect anchor events and suggest actions by analyzing user interactions, using machine learning to correlate event occurrences with action types and contextual conditions, enabling intelligent action suggestions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of operation

If a digital assistant analyzes user interactions to provide intelligent action suggestions, then user experience is improved, but computational complexity and processing time increase

Engineering Contradiction:
Improveuser experienceVSAvoidcomputational complexity
Core Design Contradiction:
Ease of operationVSDevice complexity

Solution Approach 1:

The system performs preliminary analysis by pre-processing user interaction data and pre-computing action suggestions during idle periods or background processing. The predictive model is trained in advance on historical user behavior patterns, so that when a user interaction occurs, the system can quickly retrieve and present pre-computed suggestions rather than performing complex analysis in real-time.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system applies different levels of analysis to different types of user interactions based on their importance and frequency. Common, routine interactions receive simpler, pre-computed suggestions, while rare or complex interactions trigger more sophisticated analysis. The predictive model focuses computational resources on analyzing specific event types and contexts where they provide the most value.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If the predictive model analyzes multiple contextual conditions to improve suggestion accuracy, then measurement precision is improved, but processing time increases

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

Solution Approach 1:

The system analyzes only the most relevant contextual conditions for each specific user interaction type, rather than evaluating all possible context factors. The predictive model identifies and focuses on the subset of contextual features that have the highest predictive value for suggesting actions, eliminating unnecessary analysis of less important contextual parameters.

Inventive Principle:
Principle #16Partial or excessive action

Solution Approach 2:

The system uses historical user behavior patterns as training data to create a predictive model that captures typical response patterns. Instead of analyzing all contextual conditions from scratch for each new interaction, the system copies and adapts patterns from similar historical situations, significantly reducing the computational burden while maintaining high accuracy.

Inventive Principle:
Principle #26Copying

3Productivity

If the system provides timely action suggestions based on user behavior patterns, then productivity is improved, but data processing requirements increase

Engineering Contradiction:
Improveinteraction efficiencyVSAvoiddata processing requirements
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

The system extracts and focuses on the most critical features and patterns from large volumes of user interaction data. Rather than processing and storing all raw interaction data, the predictive model identifies and extracts key behavioral patterns, event sequences, and contextual features that are most predictive of user actions, reducing the effective data volume that needs to be processed while maintaining predictive accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4400966B1Suggesting executable actions in response to detecting events
Publication Date: 2025.08.27 APPLE INC
  • EP4400966B1 patent drawingFigure 1
  • EP4400966B1 patent drawingFigure 2A
  • EP4400966B1 patent drawingFigure 2B

AI summary

Systems and processes for providing, via an electronic device, suggested user actions. The suggested actions are provided in response to detecting an occurrence of a predefined event occurring in the user's day. The occurrence of the anchor is encoded in signals generated by the electronic device. The occurrence of the anchor is detectable via monitoring and analysis of electronic signals. Based on the user's previous interactions with the device, the occurrence of the anchor is indicative of user behavior and/or action taken in response to the anchor. Machine learning (ML) is employed to train an anchor model to associate actions taken in response to anchor occurrences. The trained anchor model is employed to detect anchors and provide suggested actions in response to the detected anchor occurrence. The suggested action is based on a type of anchor occurrence and contextual conditions of the anchor occurrences.