Document Anonymization via Entity Classification and Propagation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Document anonymization is challenging due to its context-sensitive nature, requiring manual labor to differentiate between entities that should be retained or anonymized, especially when linkages and indirect identifications are involved, necessitating automated assistance for efficient anonymization processes.
Innovation Solution
A document anonymization method that identifies named entities, classifies them as anonymous or public through syntactic and lexical analysis, and propagates classifications to ensure appropriate anonymization, using a tagger, default classifier, selective re-classifier, lexical processor, and syntactical processor to handle context-specific information effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual anonymization is used to ensure accurate removal of personally identifying information, then anonymization accuracy is improved, but labor intensity and time consumption increase significantly
Solution Approach 1:
The patent introduces an automated anonymization system that acts as an intermediary between the document and the final anonymized output. This system uses named entity recognition, classification, and contextual analysis components to automatically identify and anonymize personally identifying information while preserving necessary information, thereby maintaining high accuracy without requiring manual intervention for every document.
Solution Approach 2:
The patent replaces the manual mechanical process of anonymization with an automated computational system. The system uses natural language processing, syntactic analysis, and machine learning algorithms to perform tasks that previously required human judgment and manual editing, thus dramatically improving processing efficiency while maintaining anonymization accuracy through automated decision-making rules.
2Productivity
If automated anonymization is implemented to improve processing efficiency, then productivity is improved, but anonymization accuracy and reliability deteriorate due to context-sensitive challenges
Solution Approach 1:
The patent divides the anonymization process into distinct modular components: named entity recognition, entity classification, contextual analysis, and anonymization execution. Each component handles a specific aspect of the task, allowing the system to process documents efficiently while maintaining reliability through specialized processing at each stage. This segmentation enables parallel processing and reduces the complexity of any single decision point.
Solution Approach 2:
The system incorporates feedback mechanisms where the contextual analysis component continuously refines entity classifications based on surrounding text patterns and relationships. The system learns from contextual cues and adjusts its anonymization decisions accordingly, improving reliability by incorporating feedback loops that verify and refine anonymization choices before final output.
3Measurement precision
If comprehensive contextual analysis is performed to handle linkages between entities, then anonymization accuracy is improved, but device complexity and computational requirements increase
Solution Approach 1:
The patent performs preliminary named entity recognition and classification before conducting detailed contextual analysis. By pre-identifying and categorizing entities in advance, the system reduces the complexity of subsequent contextual analysis, as it only needs to evaluate relationships among already-identified entities rather than analyzing the entire document from scratch. This preliminary action streamlines the overall process while maintaining accuracy.
Solution Approach 2:
The system applies different levels of contextual analysis to different entities based on their classification and local context. Rather than performing uniform comprehensive analysis on all entities, the system focuses computational resources on entities that require deeper contextual evaluation, using local quality variations to optimize the balance between accuracy and system complexity.
Data Source
AI summary
Named entities in a document are identified. Each named entity is classified as either anonymous or public based on analysis including at least syntactic analysis of one or more portions of the document containing the named entity. In one suitable approach, each named person entity is classified by default as anonymous, and each named entity that is not a named person is classified by default as public. Named entities are selectively re-classified based on evidence contained in the document indicating that the default classification is incorrect. The classification of a named entity as either anonymous or public is propagated to multiple occurrences of that named entity in the document Those named entities classified as anonymous are anonymized.