Homogeneous Entity Grouping for Medical Knowledge Bases
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing entity grouping methods in knowledge bases, particularly in character-based languages like Chinese and Korean, face challenges in effectively grouping multiple aliases or synonyms of medical symptoms, leading to inefficient representation in medical knowledge bases due to the lack of advanced dynamic updating and natural setting adaptation.
Innovation Solution
A homogeneous entity grouping system comprising an entity feature constructor, relation predictor, criteria checker, grouper, and centroid selector, which uses similarity scores, iterative semi-supervised predictions, and human verification to form and refine entity groups, leveraging large-scale text data and prior knowledge to identify and validate synonymous entities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional entity grouping methods are used, then the knowledge base can be constructed with basic entity identification, but the grouping performance is insufficient for handling multiple aliases and synonyms in character-based languages
Solution Approach 1:
The entity grouping process is segmented into multiple independent modules: entity feature constructor, relation predictor, criteria checker, grouper, and centroid selector. Each module handles a specific aspect of the grouping task, allowing the system to achieve high accuracy through specialized processing at each stage without overwhelming complexity in a single monolithic system.
Solution Approach 2:
The system implements dynamic entity grouping that adapts to different languages and contexts. The feature constructor dynamically selects relevant features based on the input entities, and the iterative refinement process dynamically adjusts grouping decisions based on predicted relations and criterion evaluations, enabling the system to handle the variability of aliases and synonyms in character-based languages effectively.
2Measurement precision
If entity grouping is performed without iterative refinement, then the process is faster and simpler, but the grouping performance and accuracy are insufficient
Solution Approach 1:
The system performs partial iterative refinement by executing the relation prediction and criteria checking steps a fixed number of times (e.g., 5 iterations) rather than continuing until convergence. This partial action approach achieves sufficient accuracy improvement without the excessive time cost of full iterative convergence, balancing precision and efficiency.
Solution Approach 2:
The entity feature constructor performs preliminary action by pre-processing and extracting features from entities before the main grouping process begins. This preliminary feature extraction prepares the data in advance, reducing the computational burden during iterative refinement and allowing the system to achieve high accuracy more efficiently.
3Adaptability or versatility
If homogeneous entity grouping is implemented for medical symptoms, then the representation in knowledge base is improved, but the system complexity increases due to multiple processing stages
Solution Approach 1:
The entity grouping system is designed as a universal framework that can handle different types of entities (medical symptoms, diseases, procedures) and different languages (Chinese, Korean, English) through the same core modules. The feature constructor and relation predictor are designed to work universally across different entity types and languages, improving knowledge base representation quality without requiring separate specialized systems for each case.
Data Source
AI summary
Systems and methods are disclosed to identify entities that have a similar meaning, and may, in embodiments, be grouped into entity groups for knowledge base construction. In embodiments, the entity relations of similarity or non-similarity for an entity pair are predicted as a binary relationship. In embodiments, the prediction may be based upon similarity score between the entities and the entity features, which features are constructed using an entity feature or representation model. In embodiments, the prediction may be an iterative process involving minimum human checking and existing knowledge update. In embodiments, one or more entity groups are formed using graph search from the predicted entity pairs. In embodiments, a group centroid entity may be selected to represent each group based on one or more factors, such as its generality or popularity.