The Future of the UMLS Semantic Network


A workshop organized on April 7-8, 2005
at the U.S. National Library of Medicine (NLM) in Bethesda, Maryland.







The Semantic Network is one of three knowledge sources developed and distributed by the National Library of Medicine as part of the Unified Medical Language System® (UMLS®) project, along with the Metathesaurus® and the SPECIALIST lexicon. The Semantic Network was created in an effort to provide a semantic framework for the UMLS and its constituent vocabularies. Unlike the Metathesaurus, the Semantic Network is a small structure composed of 135 high-level categories called semantic types. It is organized in two single-inheritance hierarchies: one for Entity and one for Event. In addition to is a, 53 kinds of relationships are defined in the Semantic Network, which are used to represent over 6,700 relations – hierarchical and associative – among semantic types. Semantic types from the Semantic Network are linked to Metathesaurus concepts by the categorization link established by the Metathesaurus editors: Each concept is categorized with at least one semantic type from the Semantic Network, independently of its hierarchical position in the source vocabularies. Fifteen collections of semantic types, called semantic groups, have been defined in order to partition Metathesaurus concepts in a smaller number of semantically consistent groups.


Reported applications of the Semantic Network include natural language processing (NLP) and text mining (e.g., to suggest or validate relations among entities identified in text), information retrieval and navigation (e.g., to constrain search or display of UMLS concepts with respect to semantic types) and alignment and interoperability (e.g., to help find similarity among concepts based on similarity among high-level categories). Additionally, the categorization of Metathesaurus concepts allows users to select all Metathesaurus concepts corresponding to a given category (e.g. cancers, through the semantic type Neoplastic Process). More generally, the overarching structure provided by Semantic Network enables the analysis of differences between relations asserted among concepts in the Metathesaurus and among the corresponding semantic types in the Semantic Network.


Unlike the other two knowledge sources, the Semantic Network has been essentially stable over the last ten years. However, some of its users have suggested changes regarding coverage (additional semantic types and relations), content (update to reflect current biomedical science), organization (support for multiple inheritance) and compatibility with other ontologies (compliance with ontological principles). Although less crucial to its mission, changes in formalism (e.g., description logics) have also been suggested.


NLM is holding a workshop designed to foster discussions about possible evolution of the UMLS Semantic Network. The goals are: to identify current limitations of the Semantic Network and their impact on applications and maintenance of the network itself and to analyze and discuss suggestions for changing and improving the Semantic Network.




The Semantic Network Workshop will be held in the Natcher Building (building 45) on the National Institutes of Health campus. Please refer to the visitor information section for maps and directions. The following web site provides a list of hotels convenient to the NIH: .






Thursday, April 7, 2005 (Natcher Building, Room A)





Betsy L. Humphreys, Deputy Director, NLM



The UMLS Semantic Network



Alexa T. McCray, Harvard University

From birth to the present


Reported uses



Kin Wah Fung, NLM

UMLS questionnaire



Yehoshua Perl, New Jersey Institute of Technology

NJIT questionnaire


Coffee break


Structural issues



Yehoshua Perl, New Jersey Institute of Technology

Enriching and designing metaschemas for the UMLS Semantic Network



James Geller, New Jersey Institute of Technology

The refined Semantic Network




Ontological issues



Anand Kumar, IFOMIS, Germany

Ontological analysis of the semantic types



Barry Smith, University at Buffalo

Semantic Network and biomedical ontology


Limitations for natural language processing


Thomas C. Rindflesch, NLM

The UMLS Semantic Network for semantic interpretation


Coffee break


Limitations for semantic integration


Anita Burgun, University of Rennes, France

The UMLS Semantic Network: Support for semantic integration and reasoning



Friday, April 8, 2005 (Natcher Building, Room A)



Limitations for uses in clinical systems


James J. Cimino, Columbia University

Experience with using the UMLS Semantic Network to coordinate controlled terminologies for a large clinical data repository


Limitations for UMLS maintenance


Olivier Bodenreider, NLM

Consistency between Metathesaurus and Semantic Network


Coffee break


Stuart J. Nelson, NLM

Semantic Network complexity and Metathesaurus maintenance




A summary of issues and suggestions


Olivier Bodenreider, NLM



Discussion: Possible evolutions


Moderator: Betsy L. Humphreys


Panel discussion: Commentary on proposed evolution

Olivier Bodenreider, Anita Burgun, James J. Cimino, Alexa T. McCray, Stuart J. Nelson, Thomas C. Rindflesch, Barry Smith


Suggestions and questions from the audience





Visitor information


Natcher Conference Center

NIH Campus, Bldg. 45


The Natcher Conference Center is located in the Natcher Building on the NIH campus in Bethesda, Maryland adjacent to the U.S. National Library of Medicine. All visitors must show one form of government issued photo identification to get into Natcher (e.g., driver’s license, passport, green card, etc.) Visitors should be prepared to go through a metal detector and have their personal belongings inspected.  Visitors will be required to wear a visitor’s pass at all times while attending the workshop.  Participants are encouraged to leave extra bags or personal materials at their hotel to minimize the inspection time.



Participants are encouraged to take public transportation to the NIH campus. The Natcher Building is located on Center Drive and is a 5-minute walk from the Metro Medical Center Station on the Red Line.


By Subway

Take the Red Line and exit at the Medical Center Station. At the top of the station's escalators, take the stairs or ramp to your left and follow the path to the Natcher Building.  For additional information about the Metro subway system go to the Metro website:



By Air

The Washington, DC area is served by all three major airports (e.g., Ronald Reagan National, Washington-Dulles International, and Baltimore-Washington International). Travel time from all three airports to Bethesda area hotels and the NIH campus is approximately 45 minutes to 1 hour.


By Train

Union Station serves the Washington, DC area. It is approximately 45 minutes to Bethesda area hotels and the NIH campus from the station.


By Taxi

From Ronald Reagan National Airport: 12 miles from NIH; one-way fare is approximately $35.

From Dulles International Airport: 26.2 miles from NIH; one-way fare is approximately $45.

From Baltimore-Washington Airport: 34.5 miles from NIH; one-way fare is approximately $55. Various Airport shuttles can take you into Bethesda for approximately $30. 



All vehicles will be inspected. There are two entrances for visitors who are driving to the campus:  Center Drive from Old Georgetown Road (Rte 187) and South Drive from Wisconsin Avenue.  Visitors must show one form of government issued photo identification to get onto the NIH Campus (e.g., driver’s license, passport, green card, etc.) There are a limited number of visitor parking spaces available at the Natcher building. Alternatively, participants may wish to park in a public lot in Bethesda and walk to the NIH campus. Participants may also drive and park at a Metro station that has public parking and then take the Metro to the Medical Center station. Convenient Metro stops with parking include Rockville, Twinbrook, White Flint, and Grosvenor.  There is not parking available at the Medical Center. 


A map of the NIH campus and general information about parking and driving to campus can be found at: