UMLS (Unified Medical Language System)

Roy Williams roy at cacr.caltech.edu
Mon Sep 30 10:44:51 PDT 2002


> Amarnath Gupta at SDSC has implemented the UMLS ontology ....
> We have used the ontology for tasks such as:
> - parse a report to identify relevant words
> - use the UMLS ontology to identify a broader context that includes
> the identified words
> - access the medpub collection to retrieve similar reports on the same
subject.

Reagan and Amarnath

This sounds like an interesting application of an ontology. I have been
thinking of a similar type of project, taking a paper about astronomy,
finding all the references to galaxies and stars (eg NCG 666), and adding
the relevant citations to NED and Simbad.

A couple of questions:

-- When you parse a report for relevant words, did you start with text, or
pdf, or Word, or Latex, or what?

-- Did the parsing involve a Natural Language parser, or was it a tokenizing
on whitespace?

-- The UMLS has a lot of structure besides a simple set of words. Were any
parts of this structure used?

Roy



More information about the semantics mailing list