Feature Value Conversion for Semantic Content Search

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing technologies face challenges in generating feature values that accurately represent semantic factors for content such as images, videos, and sounds, especially when manual labeling is labor-intensive and sufficient pairs of related content are not available, limiting their applicability and precision in finding semantically similar content.

Innovation Solution

A method and device for learning feature value conversion functions that generate low-dimensional feature values by selecting content pairs with high correlation and iteratively refining transformation matrices to maximize the correlation between low-dimensional feature values, even with moderate relevance between content groups, allowing for the identification of semantically similar content without requiring extensive manual labeling or paired content.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual labeling is used to generate semantic labels for images, then semantic accuracy is improved, but labor cost and time consumption increase significantly

Engineering Contradiction:

Improvesemantic accuracyVSAvoidlabeling time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs automatic semantic labeling by learning feature value conversion functions from paired content groups (images and documents) without requiring manual human intervention. The computer automatically extracts initial feature values, selects content pairs, learns conversion functions, and generates semantic labels, making the system self-sufficient and eliminating the need for labor-intensive manual labeling while maintaining high semantic accuracy

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the mechanical manual labeling process with an automated computational system. Instead of human operators manually assigning semantic labels to images, the system uses computer-based feature extraction, content pair selection, and machine learning algorithms to automatically generate semantic labels, substituting human mechanical work with automated computational processes

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Ease of operation

If feature values indicating physical nature are used, then measurement is simplified, but semantic relatedness cannot be correctly evaluated

Engineering Contradiction:

Improvemeasurement simplicityVSAvoidsemantic evaluation accuracy

Core Design Contradiction:

Ease of operationVSMeasurement precision

Solution Approach 1:

The patent transforms physical feature values into semantic feature values by learning feature value conversion functions. Instead of directly using raw physical measurements (color histograms, frequency characteristics), the system learns conversion functions that map these physical features to semantic representations that correctly evaluate semantic relatedness, changing the parameter representation from physical to semantic domain

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces feature value conversion functions as an intermediary between physical feature values and semantic evaluation. These conversion functions act as a mediator that transforms raw physical measurements into semantic representations, enabling the system to bridge the gap between simple physical measurement and accurate semantic evaluation without directly comparing physical features

Inventive Principle:

Principle #24Intermediary (Mediator)

3Measurement precision

If vast numbers of content pairs are collected for learning, then learning precision is improved, but data collection complexity and storage requirements increase

Engineering Contradiction:

Improvelearning precisionVSAvoiddata collection complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent performs preliminary content pair selection by computing similarity between content groups and selecting only the most relevant pairs for learning. Instead of collecting and processing all possible content pairs, the system pre-selects a subset of high-similarity pairs that are most useful for learning feature value conversion functions, reducing data collection complexity while maintaining learning precision

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11615132B2Feature amount generation method, feature amount generation device, and feature amount generation program

Publication Date: 2023.03.28 NIPPON TELEGRAPH & TELEPHONE CORP

AI summary

Low-dimensional feature values with which semantic factors of content are ascertained are generated from relevance between sets of two types of content.Based on a relation indicator indicating a pair of groups indicating which groups are related to first types of content groups among second types of content groups, an initial feature value extracting unit 11 extracts initial feature values of the first type of content and the second type of content. A content pair selecting unit 12 selects a content pair by selecting one first type of content and one second type of content from each pair of groups indicated by the relation indicator. A feature value conversion function generating unit 13 generates feature conversion functions 31 of converting the initial feature values into low-dimensional feature values based on the content pair selected from each pair of groups.