Entity Disambiguation via Knowledge Base Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing natural language processing (NLP) methods for entity disambiguation, such as entity linking and named entity recognition, rely heavily on contextual analysis, which requires extensive computational resources and is not scalable for large volumes of data, making them inefficient for annotating large collections or short texts.
Innovation Solution
A method that uses a two-step process to disambiguate named entities in unstructured text by modifying them based on predefined rules to generate candidate entities, which are then matched against a knowledge base like Wikipedia, without relying on textual context, allowing for efficient offline annotation and hyperlinking to corresponding articles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contextual analysis is used for entity disambiguation, then disambiguation accuracy is improved, but computational resource consumption increases and scalability deteriorates
Solution Approach 1:
The patent segments the entity disambiguation process into two independent phases: (1) entity spotting/extraction from text, and (2) entity disambiguation via knowledge base matching. This segmentation eliminates the need for contextual analysis while maintaining disambiguation accuracy, thereby improving scalability and reducing computational resource consumption.
Solution Approach 2:
The patent extracts and removes the contextual analysis component from the entity disambiguation process. By taking out this computationally expensive element and replacing it with direct knowledge base matching, the system achieves comparable accuracy with significantly reduced computational requirements, enabling better scalability.
2Measurement precision
If contextual analysis is used for entity disambiguation, then disambiguation accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments the entity disambiguation process into two independent phases: (1) entity spotting/extraction from text, and (2) entity disambiguation via knowledge base matching. This segmentation eliminates the need for contextual analysis while maintaining disambiguation accuracy, thereby improving scalability and reducing computational resource consumption.
Solution Approach 2:
The patent performs preliminary entity spotting and extraction before disambiguation, creating a simplified input for the matching phase. This preliminary action eliminates the need for time-consuming contextual analysis during the disambiguation process itself, reducing overall processing time while maintaining accuracy.
3Measurement precision
If rule-based modification is applied to generate candidate entities, then recall is improved, but processing complexity increases
Solution Approach 1:
The patent applies parameter changes to entity names through a set of predefined modification rules (e.g., capitalization changes, plural/singular transformations, article removal). These parameter changes generate candidate entities that improve recall by matching various forms of the same entity in the knowledge base, while the rules remain simple and computationally efficient.
Solution Approach 2:
The patent creates multiple copies of the input entity with different modifications (capitalization, plurality, articles) to generate candidate entities. This copying approach with simple transformations improves recall by covering various entity name variations without requiring complex processing logic.
Data Source
AI summary
A method comprising: operating at least one hardware processor for: receiving, as input, at least one named entity, modifying said named entity based on a plurality of modification rules to generate a set of candidate named entities corresponding to said named entity, and identifying, for at least one candidate named entity in said set of candidate named entities, an article in a knowledge base of articles, wherein a title of said article matches said candidate named entity.