Semi-Automatically Inducing Semantic Classes of Clinical Research Eligibility Criteria Using UMLS and Hierarchical Clustering

This paper presents a novel approach to learning semantic classes of clinical research eligibility criteria. It uses the UMLS Semantic Types to represent semantic features and the Hierarchical Clustering method to group similar eligibility criteria. By establishing a gold standard using two independ...

Full description

Saved in:
Bibliographic Details
Published in:AMIA ... Annual Symposium proceedings 2010-11, Vol.2010, p.487-491
Main Authors: Luo, Zhihui, Johnson, Stephen B, Weng, Chunhua
Format: Article
Language:English
Subjects:
Biomedical Research
Cluster Analysis
Humans
Models, Theoretical
Natural Language Processing
Semantics
Unified Medical Language System
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:This paper presents a novel approach to learning semantic classes of clinical research eligibility criteria. It uses the UMLS Semantic Types to represent semantic features and the Hierarchical Clustering method to group similar eligibility criteria. By establishing a gold standard using two independent raters, we evaluated the coverage and accuracy of the induced semantic classes. On 2,718 random eligibility criteria sentences, the inter-rater classification agreement was 85.73%. In a 10-fold validation test, the average Precision, Recall and F-score of the classification results of a decision-tree classifier were 87.8%, 88.0%, and 87.7% respectively. Our induced classes well aligned with 16 out of 17 eligibility criteria classes defined by the BRIDGE model. We discuss the potential of this method and our future work.
ISSN:1559-4076