We describe one approach for natural language processing and database
representation of medical information. The method is based in the sema
ntic analysis of the statements and in the identification of patterns.
The processed information is indexed and structured into a frame form
at containing semantic slots into the database. We tested our method a
nalysing sentences describing symptoms extracted from case reports pre
sented in the volume 328 of the New England Journal of Medicine. The r
esults are: 73.41% of the sentences were formatted; 81.05% of the anal
ysed words were identified; and 95.33% of the medical terms were index
ed. We conclude that this semantic approach is not only efficient for
processing natural language texts, but it can also be used for the org
anization of medical information using a language-independent format.
This interlingua that is set into the database can be applied to seman
tic data retrieval; serving as a basis to organize a problem-orientate
d medical record; displaying simultaneously the DB information into tw
o or more languages; information interchanging among different human l
anguages and computers; and automatic translation, among others.