"Expertness" from Structured Text? RECONSIDER: A Diagnostic Prompting Program

" p a r s e d ''~ to p r oduc e a s e q u e n c e of ts~w.~, each with the following a t t r ibu tes : ordinal posi t ion of t e em in p h r a s e ordinal posi t ion of p h r a s e in clause ordinal posi t ion of c lause in s e n t e n c e ordinal posi t ion of s e n t e n c e in p a r t n a m e of p a r t d i sease body s y s t e m ( s ) of d isease Thus. a d ic t ionary (conta in ing in excess of 20.000 such t e r m s ) was fo rmed and CM.IT "inverted", so that each dictionary entry was folLowed by pointers to every occurrence of that en t ry in CMIT. Included with every po in te r were the seven attributes associated with each o c c u r r e n c e of that t e rm. There a re 333.211 t e r r a o c c u r r e n c e s in CMIT. for an average of a b o u t 102 t e r m s p e r disease, or 79 unique t e r m s p e r disease, the difference being t e r m s tha t are u sed m o r e t han once in a given deflnitior~ In principle, this "dic t ionary" could be used to r e c o n s t r u c t CMIT, as i t r~preswrtts, i n aZter,tat~ve for.7.at, e=actty tlw saw~ ~nformat~rn! This Large inver ted file allows et~cient searching for terms in the text. The searches can be (I) constrained to a context (diseases of the skin). (2) c o n s t r a i n e d to t ex tua l p rox imi ty (adjacency. or membership within a clause), or (3) constrained to a definition p a r t (symptoms