Knowledge Graph Entity Reconciliation via Subgraph Similarity

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing knowledge graphs face challenges in efficiently and reliably incorporating internet data sources due to duplication and erroneous additions, necessitating manual maintenance to ensure accuracy and relevance.

Innovation Solution

A computer-implemented method that crawls the internet to locate data sources, parses entities and relationships, generates sub-graphs, and compares them with existing knowledge graphs to determine similarity, merging or discarding data based on similarity scores, with manual processing for non-matching sources.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If manual maintenance is used to ensure accuracy, then data quality is improved, but productivity deteriorates

Engineering Contradiction:

Improvedata qualityVSAvoidmaintenance efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The system performs self-service by automatically crawling internet data sources, parsing entities and relationships, generating sub-graphs, and comparing them with existing knowledge graphs to determine whether to merge or discard data, thereby maintaining data quality without requiring manual intervention for each data integration task

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the manual mechanical process of knowledge graph maintenance with an automated computational system that uses web crawling, natural language processing, graph theory algorithms, and machine learning models to perform data extraction, validation, and integration tasks

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Quantity of substance

If internet data sources are used to extend knowledge graph, then quantity of data is improved, but reliability deteriorates due to duplication and erroneous additions

Engineering Contradiction:

Improvedata volumeVSAvoiddata accuracy

Core Design Contradiction:

Quantity of substanceVSReliability

Solution Approach 1:

The system introduces an intermediary validation layer between internet data sources and the knowledge graph, consisting of sub-graph generation, similarity comparison algorithms, and confidence scoring mechanisms that filter and validate incoming data before integration, preventing duplication and erroneous additions

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system implements feedback loops where extracted entities and relationships are compared against existing knowledge graph data, with similarity scores and confidence metrics providing feedback to determine whether to accept, reject, or flag data for manual review, ensuring data accuracy while maximizing data volume integration

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS10423652B2Knowledge graph entity reconciler

Publication Date: 2019.09.24 BAIDU USA LLC

AI summary

Systems and methods are disclosed for extending and reinforcing a knowledge graph using wiki-like web pages as a source of information. A web crawler parse a wiki-like source and obtain a topic entity from the source. Relationships between the topic entity and sub-topics within the source are identified and a graph is built with the topic and relationships to sub-topics. A candidate topic in the knowledge graph is identified, and a sub-graph of the knowledge graph is either identified or generated. The knowledge sub-graph contains the candidate topic and relationships to sub-topics. A similarity is computed between the source graph and the knowledge sub-graph. If the two graphs are sufficiently similar, then the source topic graph is merged with the knowledge graph.