Date of Award

Spring 2004

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computing Sciences - (Ph.D.)

Department

Computer Science

First Advisor

Yehoshua Perl

Second Advisor

James Geller

Third Advisor

James A. McHugh

Fourth Advisor

James J. Cimino

Fifth Advisor

Michael Halper

Sixth Advisor

Huanying Gu

Abstract

The disparate terminologies used by various biomedical applications or professionals make the communication between them more difficult. The Unified Medical Language System (UMLS) of the National Library of Medicine (NLM) is an attempt to integrate different medical terminologies into a unified representation framework to improve decision making and the quality of patient care as well as research in the health-care field. Metathesaurus (META) and Semantic Network (SN) are two main components of the UMLS system, where the SN provides a high-level abstract of the concepts in the META.

This dissertation addresses three problems of the SN. First, the SN's two-tree structure is restrictive because it does not allow a semantic type to be a specialization of several other semantic types. This restriction leads to the omission of some subsumption knowledge in the SN. Secondly, the SN is large and complex for comprehension purposes and it does not come with a pictorial representation for users. As a partial solution for this problem, several metaschemas were previously built as higher-level abstractions for the SN to help users' orientation. Third, there is no efficient method to evaluate each metaschema. There is no technique to obtain a consolidated metaschema acceptable for a majority of the UMLS's users.

In this dissertation work the author attacked the described problems by using the following approaches. (1) The SN was expanded into the Enriched Semantic Network (ESN), a multiple subsumption structure with a directed acyclic graph (DAG) IS-A hierarchy, allowing a semantic type to have multiple parents. New viable IS-A links were added as warranted. Two methodologies were presented to identify and add new viable IS-A links. The ESN serves as an extended high-level abstract of the META. (2) The ESN's semantic relationship distribution and concept configuration were studied. Rules were defined to derive the ESN's semantic relationship distribution from the current SN's semantic relationship distribution. A mapping function was defined to map the SN's concept configuration to the ESN's concept configuration, avoiding redundant classifications in the ESN's concept configuration. (3) Several new metaschemas for the SN and the ESN were built and evaluated based on several different partitioning techniques. Each of these metaschema can serve as a higher-level abstraction of the SN (or the ESN).

Share

COinS