Shared Learning Model for Product Attribute Prediction Load Reduction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing e-commerce systems face a significant processing load due to the large number of products and attributes, making it inefficient to predict various attributes related to products.

Innovation Solution

An information processing apparatus and method using a shared learning model with multiple estimation layers to predict attributes, where the model includes an input layer for feature extraction and an output layer for concatenating attribute values, reducing processing load by predicting only valid attributes based on the product type.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of information

If preset attributes are stored for all products, then attribute information completeness is improved, but processing load increases

Engineering Contradiction:
Improveattribute information completenessVSAvoidprocessing load
Core Design Contradiction:
Loss of informationVSProductivity

Solution Approach 1:

The patent replaces the mechanical approach of storing preset attributes with an image-based prediction system using deep learning models. Instead of manually storing and managing attribute data for each product, the system uses neural networks to automatically predict attributes from product images, thereby reducing processing load while maintaining information completeness.

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

Solution Approach 2:

The patent changes the fundamental parameter of attribute representation from stored data to predicted data. By transforming the system from storing predefined attributes to dynamically predicting attributes based on image analysis, the processing load is reduced while maintaining comprehensive attribute information through the prediction model.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If multiple attributes are predicted for each product, then attribute coverage is improved, but prediction complexity increases

Engineering Contradiction:
Improveattribute coverageVSAvoidprediction complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the attribute prediction process into multiple specialized estimation layers, where each layer is responsible for predicting specific types of attributes. This segmentation allows the system to handle diverse attribute coverage while managing complexity by dividing the prediction task into modular, specialized components rather than a single monolithic prediction system.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent creates a universal prediction model that can handle multiple attribute types across different product categories. The deep learning model is designed to be multi-functional, capable of predicting various attributes (color, size, material, etc.) for different product types using a unified architecture, thereby achieving broad attribute coverage without proportionally increasing overall system complexity.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4220551B1Information processing device, information processing method, and program
Publication Date: 2026.04.22 RAKUTEN GROUP INC
  • EP4220551B1 patent drawingFigure 1~2A
  • EP4220551B1 patent drawingFigure 2B~3A
  • EP4220551B1 patent drawingFigure 3B

AI summary

An information processing apparatus (1) includes: an acquisition means (101) for acquiring an object image including an object; and a prediction means (102) for predicting various attributes relating to the object by applying a learning model to the object image acquired by the acquisition means, wherein the learning model is a learning model that is shared between a plurality of different objects including the object and includes a plurality of estimation layers that estimate a plurality of attribute values for a plurality of attributes relating to the plurality of different objects.