Bayesian algirithms for sorting email spam out are good. You have to
train them on your particular real correspondents and your email, but
then they are very good. Better better than people with overwhelming
amounts of spam, I suspect less good than people on the occasional one
that gets through, which we recognise almost instantly, but able to
learn from that.
http://www.joelonsoftware.com/articles/FogBugzII.html
This software developer trained a filter to sort incoming proper email
("ham") into categories.
There is a link to a paper by the summer intern he had do the
probability work, about it.
Why are we still not using Bayesian techniquest to send incoming mail to
the right doctor/nurse/administrator?
Why are we not using them to sort patients into those who deserve a
second look because they are sort of like someone who had a bad outcome
recently? (I thought of that one several years BG, when Excite first
came to my attention as a search engine).
Answer:
Because we are hidebound? No.
" " " stupid? No.
" " no longer have control of development of our software tools,
but they are instead directed by people who do not regard the domain and
domain knowledge of medicine as a proper driving force for development?
Well, maybe getting warm there on some of it.
No, mainly because the hospitals keep turning out pieces of apper, and
because IT in the NHS is becomeiong a matter of moving and looking at
_pictures of pieces of paper_ and regarded as big improvement on the
actual pieces of paper (and it is, to some extent, although the pieces
of paper seem to get moved as well, so there is double working going on
which cannot be good aargh....)
So lacking a handwriting recognition tool, and with OCR of even typed
pape and images of paper not making much progress yet, in the field, and
with the spasm of capitalism and the NHS takeover of IT provision
combining to make it harder, not easier, to develop even something as
simple as that, we do not progress on using the information.
Pity.
BG: before Google, when the world was young and the tubes not full.
--
A
