Entity Disambiguation for Person-Centric Cross-Space Information Linking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The segregation of information across different spaces (public, semi-private, and private) leads to information overload and inefficiency in accessing relevant information, as existing methods are application-centric, domain-centric, or interest-centric, failing to establish meaningful connections and requiring manual integration of data from multiple sources.
Innovation Solution
A person-centric INDEX system that cross-links and organizes information from various spaces into a unified, dynamic space tailored to the individual, using entity extraction and disambiguation to provide intent-based information presentation and automated task completion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If information is organized in segregated spaces (public, semi-private, private) using conventional application-centric or domain-centric methods, then information can be stored and accessed within each space, but information overload occurs and meaningful connections between related information across spaces are lost
Solution Approach 1:
The patent merges information from multiple segregated spaces (public, semi-private, private) into a unified person-centric knowledge graph. This integration preserves meaningful connections between related information across different spaces by representing all entities and their relationships in a single connected structure, eliminating the information loss that occurs in segregated organization.
Solution Approach 2:
The knowledge graph serves as a universal structure that handles multiple functions: storing information from diverse sources, establishing semantic relationships, enabling entity disambiguation, and supporting various query types. This multi-functional approach reduces the need for separate systems for each function, simplifying the overall information organization complexity.
2Ease of operation
If conventional application-centric or domain-centric methods are used to organize information, then each application or domain can manage its own information subset, but users must manually search and integrate information across multiple segregated spaces
Solution Approach 1:
The system performs preliminary action by automatically extracting entities, determining entity types, identifying candidate entities, and establishing relationships between entities across different spaces before user queries. This pre-processing creates a ready-to-query knowledge graph structure, eliminating the need for users to manually search and integrate information across multiple spaces at query time.
Solution Approach 2:
The knowledge graph system provides self-service by automatically organizing information from multiple spaces, resolving entity ambiguities, and establishing semantic relationships without user intervention. The system autonomously builds and maintains the connected information structure, freeing users from manual information integration tasks.
3Measurement precision
If entity names are extracted from data sources without disambiguation, then entity extraction is simple, but entity candidates cannot be accurately identified when multiple entities share the same name
Solution Approach 1:
The entity disambiguation process is segmented into distinct stages: entity name extraction, entity type determination, candidate entity identification, and final entity selection. Each stage processes specific aspects of disambiguation using appropriate methods, making the overall complex process manageable and systematic while improving identification precision.
Solution Approach 2:
Entity type determination serves as an intermediary step between simple entity name extraction and final candidate identification. By introducing entity types as a mediating classification layer, the system narrows down candidate entities and improves identification precision without requiring direct complex disambiguation logic at each step.
Data Source
AI summary
The present teaching relates to entity extraction and disambiguation. In one example, an entity name extracted from a data source associated with a user is obtained. One or more entity types associated with the entity name are determined. One or more entity candidates are identified with respect to each of the one or more entity types. An entity candidate is selected with respect to one of the one or more entity types to be an individual associated with the entity name.