Entity Disambiguation via Context Segmentation and Search Keys
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing entity disambiguation systems face challenges in accurately resolving mentions of names, words, or phrases in text to distinct persons, organizations, or locations across large datasets, especially with variations in spelling and transliteration, leading to confusion and inefficiencies in data processing.
Innovation Solution
A method and apparatus for entity disambiguation using language-independent algorithms that leverage multiple types of context evidence, automatically learning the weight of evidence from corpus statistics, and employing efficient search keys to identify pairs of mentions corresponding to the same entity, thereby reducing unnecessary comparisons and improving throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional entity disambiguation methods are used, then accuracy in resolving mentions can be improved, but computational time and processing efficiency deteriorate due to comparing all possible mention pairs
Solution Approach 1:
The patent segments the entity disambiguation task by dividing mentions into groups based on contextual features (such as document sections, paragraphs, or semantic clusters). This segmentation reduces the search space from all possible mention pairs to only pairs within segmented groups, maintaining accuracy while significantly improving processing throughput.
Solution Approach 2:
The patent performs preliminary actions by pre-computing and storing contextual information about mentions (such as surrounding text, document metadata, or entity type annotations) before the disambiguation process. This preliminary preparation enables faster comparison and reduces computational time during the actual disambiguation phase.
2Measurement precision
If all mention pairs are compared to ensure comprehensive disambiguation, then resolution accuracy improves, but the number of unnecessary comparisons increases
Solution Approach 1:
The patent applies local quality by evaluating mentions based on their specific contextual characteristics rather than treating all mentions uniformly. By assigning different weights and criteria to mentions based on their local context (such as proximity, semantic similarity, or document role), the system focuses computational resources on the most relevant comparisons, reducing time spent on unnecessary comparisons while maintaining high resolution accuracy.
3Measurement precision
If multiple context types are used for disambiguation, then disambiguation performance improves, but system complexity increases
Solution Approach 1:
The patent implements universality by designing a unified disambiguation framework that can handle multiple context types (textual, structural, semantic) through a single integrated mechanism. Rather than requiring separate processing systems for each context type, the framework provides a universal approach that processes all context types through consistent algorithms, improving performance while managing system complexity through standardization.
Data Source
AI summary
Entity disambiguation resolves which names, words, or phrases in text correspond to distinct persons, organizations, locations, or other entities in the context of an entire corpus. The invention is based largely on language-independent algorithms. Thus, it is applicable not only to unstructured text from arbitrary human languages, but also to semi-structured data, such as citation databases and the disambiguation of named entities mentioned in wire transfer transaction records for the purpose of detecting money-laundering activity. The system uses multiple types of context as evidence for determining whether two mentions correspond to the same entity and it automatically learns the weight of evidence of each context item via corpus statistics. The invention uses multiple search keys to efficiently find pairs of mentions that correspond to the same entity, while skipping billions of unnecessary comparisons, yielding a system with very high throughput that can be applied to truly massive data.