home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
ftp.cs.arizona.edu
/
ftp.cs.arizona.edu.tar
/
ftp.cs.arizona.edu
/
icon
/
newsgrp
/
group98b.txt
/
000015_icon-group-sender _Wed May 6 12:36:51 1998.msg
< prev
next >
Wrap
Internet Message Format
|
2000-09-20
|
3KB
Return-Path: <icon-group-sender>
Received: from kingfisher.CS.Arizona.EDU (kingfisher.CS.Arizona.EDU [192.12.69.239])
by baskerville.CS.Arizona.EDU (8.8.8/8.8.7) with SMTP id MAA07362
for <icon-group-addresses@baskerville.CS.Arizona.EDU>; Wed, 6 May 1998 12:36:51 -0700 (MST)
Received: by kingfisher.CS.Arizona.EDU (5.65v4.0/1.1.8.2/08Nov94-0446PM)
id AA15209; Wed, 6 May 1998 12:36:47 -0700
From: gep2@computek.net
Date: Wed, 6 May 1998 12:04:42 -0500
Message-Id: <199805061704.MAA14525@axp.cmpu.net>
Mime-Version: 1.0
Content-Type: text/plain
Content-Transfer-Encoding: 7bit
Subject: AI use for Icon
To: icon-group@optima.CS.Arizona.EDU
X-Mailer: SPRY Mail Version: 04.00.06.17
Errors-To: icon-group-errors@optima.CS.Arizona.EDU
Status: RO
Content-Length: 2300
> I'm writing a program with icon that reads an email message and
determines the who, what, when, where and by whom. At least that is what
I am trying to do. This message would come form a supervisor to the
human resources department asking them to promote Mr. Any Body to a New
Title on a certain date. By having all the employees names prior to
running the program I weight each name based on its reletive position to
the other name within the organization to get a fuzzy match. One of
which must be a supervisor. I've never tried using AI technology before
so I looking for someone who has done something similar that would share
with me their text parse-ing process for ideas. I'm currently looking
for words that match my list as for the 'what' the matching terms would
be 'promote, demote, reassign, realign, fire...'. Likewise, 'who' is
the list of employees.
Sounds like an interesting problem!
For the "fuzzy match" I think that one interesting way to at least help winnow
down the possibilities would be to examine the intersection of the character
sets of the different names. Those which have a high intersection (all but a
"few" characters) can be then examined more closely.
For a better "fuzzy compare" function I've liked the use of overlapping
character pairs (including a blank added to the start and end of each name).
I think that character ADJACENCY is an important aspect of the "visual
similarity" of two words or names.
> I'm now working on the 'when' to search for dates. I need a different
method than giving the program every possible date in every possible
way. Does someone have a program that would pick a date out of a
sentence?
Certainly it ought to be very easy to parse NUMERIC dates. And since most
formats of dates (with the exception of things like "the first of May") have
numeric characters associated with them, I wouldn't think it would be so
difficult to zero in on a very small number of possibilities, which could then
be processed by more exhaustive "multi-alternative" patterns.
It sounds like a fun project, in any case... if you decide you'd like to farm it
out, I would enjoy taking a stab at it. :-)
Gordon Peterson
http://www.computek.net/public/gep2/
Support the Anti-SPAM Amendment! Join at http://www.cauce.org/