Entity Disambiguation for Alert Systems
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional alert systems based on keyword matching often result in ambiguous references to named entities, leading to numerous irrelevant alerts, as they fail to distinguish between different entities with the same name, causing users to receive unwanted notifications.
Innovation Solution
A system and method for disambiguating features in alert systems, which accurately associates new information with the correct entity of interest by using entity disambiguation, allowing users to receive alerts only for newly disambiguated features such as new persons, places, or companies, and storing this information in an alert database for targeted notifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional keyword-based alert systems are used, then the system can notify users about information containing specific keywords, but it generates many false positives by failing to distinguish between different entities with the same name
Solution Approach 1:
The system performs entity disambiguation in advance by creating a knowledge base that pre-processes and resolves entity references before alert generation. This preliminary action of disambiguating entities and establishing their relationships allows the system to accurately identify which entity a keyword refers to, thereby reducing false positives without requiring complex real-time analysis during alert generation
Solution Approach 2:
The patent introduces an intermediary knowledge base that acts as a mediator between the alert query and the document corpus. This knowledge base stores disambiguated entity information and serves as an intermediate layer that enables precise entity matching, allowing the system to distinguish between different entities with the same name without directly comparing every document against every entity definition
2Productivity
If the system monitors all entities with the same name, then it captures all potential matches, but users receive unwanted notifications about entities they are not interested in
Solution Approach 1:
The system performs preliminary entity disambiguation and relationship establishment in the knowledge base before generating alerts. By pre-processing entities to identify their unique characteristics and relationships, the system can later filter alerts to only those relevant to the user's specific entity of interest, maintaining high alert coverage while eliminating unwanted notifications about unrelated entities
Solution Approach 2:
The system uses feedback from the knowledge base structure and entity relationships to refine alert generation. By analyzing the disambiguated entity information and user preferences stored in the knowledge base, the system can provide feedback mechanisms that learn from past alert interactions and improve future alert relevance, reducing false positives while maintaining comprehensive monitoring
Data Source
AI summary
The present disclosure relates to a method of alerting users regarding newly disambiguated features. More specifically, a newly disambiguated feature may pass through different filters/restrictions, such as, the known knowledge base. The disclosed known knowledge base may filter the newly disambiguated feature, comparing the newly disambiguated features to the existing features to discover a new feature of interest. Particularly, the disclosed new feature of interest may include a new person, a new phone number, a new place, a new company, among others. Finally, if there is a new feature that did not match with the existing disambiguated features in the known knowledge base, then an alert may be emitted to a user.