Sparse N-Gram Relation Extraction for EHR Text Spans
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting patient-specific data from electronic health records (EHRs) face challenges such as inefficient management, inconsistent data quality, and high computational complexity, making it difficult to accurately determine subject-health entity relationships for precision medicine applications.
Innovation Solution
A sparse tensor model architecture is used to analyze EHRs, employing N-gram counts and a simple sparse learnable layer for efficient feature extraction, with active learning and artificial labeling to reduce computational burden and improve accuracy in determining subject-health entity relationships.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional dense modeling approaches are used for patient-entity relation extraction, then measurement precision may be improved, but device complexity and computational resources increase significantly
Solution Approach 1:
The patent extracts and focuses only on the most relevant features for relation extraction by using sparse N-gram modeling. Instead of processing all possible features, the system identifies and extracts specific N-gram patterns that are most indicative of patient-entity relationships, thereby reducing model complexity while maintaining extraction accuracy.
Solution Approach 2:
The patent changes the parameter representation from dense vectors to sparse N-gram counts. This parameter transformation allows the model to work with discrete, interpretable feature counts rather than continuous dense embeddings, reducing computational burden while preserving the ability to accurately capture linguistic patterns for relation extraction.
2Measurement precision
If comprehensive EHR data analysis is performed to improve relation extraction accuracy, then measurement precision improves, but loss of time and computational resources increase
Solution Approach 1:
The patent applies partial action by analyzing only the necessary portions of EHR data required for relation extraction. Instead of processing the entire EHR record uniformly, the system selectively applies N-gram analysis to relevant text spans and sections, reducing processing time while maintaining extraction accuracy for the critical patient-entity relationships.
Solution Approach 2:
The patent segments the EHR data into manageable text spans and processes them using sparse N-gram modeling. By dividing the large EHR dataset into smaller, discrete N-gram features, the system can process the information more efficiently in parallel, reducing overall processing time while capturing comprehensive relational information.
3Measurement precision
If detailed feature extraction is applied to improve extraction accuracy, then measurement precision improves, but device complexity and computational burden increase
Solution Approach 1:
The patent uses computationally inexpensive N-gram counting methods instead of expensive deep learning feature extraction. The N-gram features are simple, discrete counts that can be calculated efficiently without requiring complex computational resources, yet they provide sufficient discriminative power for accurate patient-entity relation extraction.
Solution Approach 2:
The patent replaces complex mechanical-like deep learning processing with a simpler statistical approach based on N-gram counting. Instead of using heavy neural network architectures for feature extraction, the system substitutes this with straightforward text processing and frequency counting, significantly reducing computational complexity while maintaining extraction effectiveness.
Data Source
AI summary
Methods, systems, and software are provided for determining a relationship between a subject and a health entity. An electronic health record (EHR) for the subject is split into sections by detecting delineating section headers, and sections are subdivided into text spans. Text spans are filtered by language pattern recognition into a set of text spans having an expression related to the health entity. The natural language context of the expression in each text span in the set is evaluated to obtain a corresponding scoring representation. Scoring representations are inputted into a model comprising a plurality of parameters. The model outputs, for each text span in the set, at least a prediction that the text span is associated with the health entity. Models for determining relationships between subjects and health entities and methods for training models to determine relationships between subjects and health entities are also provided.