Latent Vector Matching for Relational Data Without Predefined Distances

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing object matching techniques fail to match objects between different data sets without predefined distances or correspondences, especially when handling relational data that includes relationships between objects.

Innovation Solution

An analysis device that estimates latent vectors for relational data to characterize its structure and matches objects between different data sets by transforming objects into latent vectors that reflect their relationships, ensuring close vectors correspond to closely related objects and similar distributions across data sets.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If supervised object matching techniques are used, then matching accuracy can be improved when distances and correspondences are defined, but the method cannot handle cases where distances are not defined or correspondences are not provided

Engineering Contradiction:

Improvematching accuracyVSAvoidapplicability to undefined distance cases

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent introduces latent vectors as an intermediary representation that bridges different data sets. Instead of directly comparing objects from different data sets without defined distances, the method transforms objects into latent vectors that capture their structural relationships. This intermediary representation enables comparison and matching even when direct distance metrics between different data sets are not defined, while still maintaining matching accuracy through the preserved relational structure.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transforms objects from their original representation into latent vector representations, changing the parameter space in which objects are compared. By mapping objects to latent vectors that encode relational structure, the method enables distance computation and matching in a transformed space where meaningful comparisons can be made even when original distance metrics are not defined between different data sets.

Inventive Principle:

Principle #35Parameter changes

2Adaptability or versatility

If unsupervised object matching techniques are used, then matching can be performed without predefined correspondences, but relational data structures cannot be handled

Engineering Contradiction:

Improveability to match without predefined correspondencesVSAvoidhandling capability of relational data

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent performs preliminary transformation of objects into latent vectors that encode their relational structures before matching. By pre-processing the data to capture relational information in the latent vector representation, the method enables subsequent matching operations to handle relational data effectively without requiring complex relational reasoning during the matching process itself.

Inventive Principle:

Principle #10Preliminary action

3Measurement precision

If objects are transformed into latent vectors to preserve relationships, then matching accuracy between relational data can be improved, but computational complexity increases

Engineering Contradiction:

Improvematching accuracy for relational dataVSAvoidcomputational complexity of vector estimation

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces complex relational reasoning mechanisms with latent vector representations. Instead of directly analyzing and comparing complex relational structures between objects, the method substitutes this with vector operations in a latent space. This substitution simplifies the computational process while preserving the essential relational information needed for accurate matching.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS11321362B2Analysis apparatus, analysis method and program

Publication Date: 2022.05.03 NIPPON TELEGRAPH & TELEPHONE CORP

AI summary

Included are a vector estimation means that estimates, in response to receiving input of a plurality of pieces of relational data each including a plurality of objects and a relationship between the objects, for each piece of relational data, a latent vector for characterizing a structure of the relational data by using the objects and the relationship included in the relational data; and a matching means that matches, for each set of first relational data and second relational data different from each other of the received pieces of relational data, a first object and a second object by using a first latent vector corresponding to the first object included in the first relational data and a second latent vector corresponding to the second object included in the second relational data. The vector estimation means estimates, when the relationship indicates a close relationship between a plurality of objects to each other, latent vectors corresponding to the objects such that the latent vectors corresponding to the respective objects have a close relationship to each other.