Harmonized Object System for ML Training Data

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

In the context of machine learning model training, existing systems face challenges in handling messages from provider objects with diverse formats, which hinders efficient data exchange and utilization for training purposes, particularly in industries like travel services where different entities use varying data exchange mechanisms.

Innovation Solution

The system harmonizes messages from provider objects by generating harmonized objects with common formats using mappings, allowing for the extraction of machine learning training data and the creation of classifiers that can be provided to machine learning models, enabling effective training across diverse data formats.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If messages from provider objects with diverse formats are used directly for machine learning training, then data exchange mechanisms can maintain their original formats, but data utilization efficiency deteriorates due to format diversity hindering training processes

Engineering Contradiction:

Improvedata exchange mechanism compatibilityVSAvoidmachine learning training efficiency

Core Design Contradiction:

Adaptability or versatilityVSProductivity

Solution Approach 1:

The patent introduces a message harmonization system that acts as an intermediary between diverse provider objects and machine learning training processes. The harmonization component receives messages in various formats from different provider objects, standardizes them into a common format, and then extracts training data. This mediator resolves the contradiction by maintaining compatibility with diverse source formats while enabling efficient uniform processing for machine learning training.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transforms the format parameter of messages from diverse states to a unified standard state through harmonization. By changing the format parameter from variable (different provider-specific formats) to constant (standardized harmonized format), the system maintains adaptability to various sources while improving training efficiency through parameter uniformity.

Inventive Principle:

Principle #35Parameter changes

2Productivity

If messages are harmonized to common formats, then machine learning training data extraction efficiency is improved, but system complexity increases due to additional harmonization processing

Engineering Contradiction:

Improvetraining data extraction efficiencyVSAvoidharmonization processing complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent applies preliminary harmonization processing to messages before they are used for machine learning training. By performing the format standardization action in advance (before training data extraction), the system improves subsequent extraction efficiency while consolidating the complexity burden into a preliminary step rather than during the critical training process.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent segments the overall processing into distinct components: message reception, harmonization/format standardization, training data extraction, and model training. This segmentation isolates the complexity of harmonization processing to a dedicated component, making the system more manageable while enabling efficient data extraction in the subsequent standardized processing stage.

Inventive Principle:

Principle #1Segmentation

3Adaptability or versatility

If multiple message formats are supported directly, then provider object compatibility is maintained, but data utilization for training deteriorates due to format variability

Engineering Contradiction:

Improveprovider object compatibilityVSAvoidtraining data quality

Core Design Contradiction:

Adaptability or versatilityVSLoss of information

Solution Approach 1:

The harmonization system serves as an intermediary that preserves information from diverse provider object formats while transforming them into a unified structure suitable for training. The mediator maintains compatibility with multiple source formats during the transformation process, ensuring no critical information is lost while enabling consistent data utilization for machine learning training.

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS11367017B2Device, system and method for training machine learning models using messages associated with provider objects

Publication Date: 2022.06.21 AMADEUS SAS

AI summary

A device, system and method for training machine learning models using messages associated with provider objects is provided. One or more computing devices: receives messages associated with provider objects representing items provided by provider systems, the messages having more than one format; stores harmonized objects corresponding to the messages, the harmonized objects generated using mappings of harmonized data of the harmonized objects to message data of the messages, the harmonized objects having common formats for a harmonized object type; extracts, from the harmonized objects, for a given machine learning model, given machine learning training data; generates, for the given machine learning model, using the given machine learning training data, at least one machine learning classifier; and provides the at least one machine learning classifier to the given machine learning model at one or more servers configured to implement the given machine learning model on received provider objects.