perm filename VANMEL.TXT[1,LMM] blob sn#120554 filedate 1974-09-18 generic text, type T, neo UTF8
From: TEITELMAN
Subject: spelling corrector
To:   MASINTER


please relay this to vanmelle.

in fact, careful reading of manual will indicate what is
happening with MYCIN and MICIN, but let me explain.
first of aal, the spelling corrector does not know
if the problem is due to extra characters, too few characters,
i.e. ones omitted, or transpositions or substitutions,
so it is trying to handle all of these. furthermore,
in the interests of efficiency it must do a linear scan
left to right. 

ok it gets to YCIN versus ICIN.  since the problem may
be a transposition, e.g. YCIN a misspelling of ICYN,
it stris off the Y, remembering it, and
now is comparing CIN versus ICIN. it is
necessary to strip off only one character at a time,
as opposed to removing both the Y and the I,
because otherwise you wouldnt catch situations
where there was one extra character in the misspelled
word but everything else is the same, e.g.
MYCIN a misspelling of MCIN.

ok, now it is comparing CIN and ICIN. This looks
like a transposition of CI and IC!. so it counts
this as one transposition, and strips off the
CI and IC, left with N and IN. now it strips of the I, and the N's agree.
so it is left with two mistakes and one  ransposition.
with fastypeflg=T, this is counted as 2, else as 3.

i am open to suggestions.

warren
-------