Ingredient-Seasoning Compatibility Prediction With Preference-Aware Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for predicting compatibility between ingredients and seasonings fail to consider individual meal preferences, leading to inaccurate predictions.

Innovation Solution

A prediction apparatus and learning method that calculates an index score based on flavor data for ingredients and seasonings, using a compatibility prediction model trained on pre-aggregated evaluations from the food industry, to output a compatibility score considering personal preferences.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If compatibility is predicted only based on ingredients and seasonings without considering meal preferences, then the prediction method is simple, but the prediction accuracy is insufficient

Engineering Contradiction:
Improveprediction accuracyVSAvoidprediction model complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The prediction model is segmented into multiple independent components: ingredient feature extraction, seasoning feature extraction, and preference feature extraction. Each component processes specific aspects separately, then their results are integrated to achieve accurate compatibility prediction while maintaining manageable complexity in each segment.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The prediction system uses a composite approach by combining multiple types of features (ingredient characteristics, seasoning characteristics, and preference data) into an integrated compatibility score. This composite model leverages the strengths of each feature type to achieve higher prediction accuracy than any single feature type alone.

Inventive Principle:
Principle #40Composite materials

2Ease of operation

If a compatibility prediction model is trained on pre-aggregated evaluation results, then the model can be applied without specialized knowledge, but the model requires substantial training data

Engineering Contradiction:
Improveease of useVSAvoidtraining data volume
Core Design Contradiction:
Ease of operationVSQuantity of substance

Solution Approach 1:

The system performs preliminary action by pre-training the compatibility prediction model using aggregated evaluation results from multiple sources before deployment. This offline training phase accumulates the necessary data processing, so that when the model is used in practice, it can provide accurate predictions without requiring users to have specialized knowledge or manually aggregate data.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4597378A1Prediction device, learning device, and learning method
Publication Date: 2025.08.06 KIKKOMAN CORP
  • EP4597378A1 patent drawingFigure 1
  • EP4597378A1 patent drawingFigure 2
  • EP4597378A1 patent drawingFigure 3~4

AI summary

Provided is a prediction apparatus including: an input section; a processing section; and an output section. The processing section extracts an index score which is inputted into a compatibility prediction model for an ingredient and a seasoning. The output section outputs a compatibility score as a predicted value of compatibility between the ingredient and the seasoning, where the compatibility score has been outputted from the compatibility prediction model by inputting the index score, which has been extracted by the processing section, into the compatibility prediction model.