Re: fuzzy string matching

Bruno DiStefano (bruno@ecf.toronto.edu)
Sat, 29 May 1999 13:01:01 +0200 (MET DST)

In article <IUH33.978$6I6.50@news2.mco>, Earl Cox <ecox@metus.com> wrote:
>Fuzzy Logic measures, to some ways, the degree to which "a" is
>representative of semantic class A ("Tall men" is a measure of the degree to
>which a height "h" is representative of the fuzzy set Tall, or the current
>rain fall is representative of Heavy rain, etc.). So a fuzzy string
>comparison function fzycmpstr(a,b) might measure the degree to which string
>"a" is representative of "b". Now, the cemetery example is OK, but is, in my
>opinion, a pathological case, since we can "see" that the two strings are
>close to each other in semantic intensionality. However, the operation
>fzycmpstr("hat","cat") or ("chair","flair") might tell us the distance
>between the two strings, but tells us nothing about their closeness. We have
>to ask ourselves what good is the fuzzy comparison "hat", "cat"? I suppose
>that depends on the application (such as a spell checker). Even taking a
>distance metric and mapping it to a fuzzy set that converts distance into
>linguistic variables (so that fzycmpstr("hat","cat") is "very close") still
>tells us little about the relationship between "hat" and "cat".
>
I agree with what you say. I mean:"I understand what
you are saying". At least, I think that I understand
what you are saying. You speak about semantics.

Let's go back to "fuzzy string matching"
The spell checker is an application. Part of a
search engine could be another. I had a case in
which I located a valuable web page when I mispelt
one word while using a search engine. The author of
the web page had made the same spelling mistake!

In general, we can say that there are three types of
error when dealing with text:
1) grammar (requires functions from string.h);
2) syntax (requires parsing a sentence and identifying words);
3) semantics (requires some AI tool; fuzzy logic?).

Fuzzy String Matching can help with (1), the spell
checker with correct version suggestion.

Your beautiful example ("this food is hot") goes
beyond the current technology in my view and
is clearly a (3).

I did not speak in the terms used by Chomsky
because he wrote his fundamental work at the
same time when Zadeh was publishing about
fuzzy logic.

The original posting of this thread, summarised at
http://www.personal.kent.edu/~jtboehm/fuzzy.html
is, in my view, interesting, because it can lend
itself to a working project for a semester course.
Extending it, can lead to some useful tool for
text processing.

>I myself have built fuzzy decision systems (actually a fuzzy Case-based
>Reasoning (CBR)Systems, see my article "A Close Shave with Occam's Razor --
>The New Face of Fuzzy Logic" in the current issue of PC/AI Magazine) with
>semantic nets linking the various objects. The edge of the net contains a
>fuzzy membership (similarity) grade, much like the probability measure on a
>Markov Chain. I then have functions that use the semantic nets along with
>the fuzzy sets defining the underlying case elements to compute fuzzy
>degrees of similarity between parameters in the system. However. . . . .
>
I will definetely look for it.

>The application of fuzzy logic and fuzzy metrics to non-numeric objects (and
>events) has long been a difficult task. One of the few practitioners who has
>successfully done this is Bill Siler with his FLOPS fuzzy expert system.
>Perhaps we can get bill to wade in on this.
>
Yes...

>Earl
>
>Will Dwinnell <76743.1740@CompuServe.COM> wrote in message
><#l6ac8Pq#GA.326@nih2naac.compuserve.com>...
>>Bruno DiStefano wrote:
>>"I feel that you have a starting point. However, you are not
>>really using the fuzzy logic in what you show. You are
>>trying to determine the "error" of a reading. Yes, I know,
>>this is not what you have in mind. However, looking from the
>>outside, this is what it looks like."
>>
>>I submit that from another perspective, this algorithm does
>>involve fuzzy logic: the presented algorithm gives a fuzzy
>>membership in the class of strings that are "like" the target
>>string.
>>
>>--
>>Will Dwinnell
>
>*************************** http://www.metus.com ***************************
>Earl D. Cox AUTHOR:
>Wandering Epistemologist "The Fuzzy Systems Handbook" (1994)
>Phrenologist-for-Hire "Fuzzy Logic for Business and Industry" (1995)
>Foreteller of the Past "Beyond Humanity: CyberEvolution and Future
>Minds"
>(919) 859-1736 (vox) (1996, with Greg Paul, Paleontologist/Artist)
>(919) 851-3525 (fax) "The Fuzzy Systems Handbook, 2nd Ed." (1998)
> "Fuzzy Tools for Data Mining" (due Summer, 1999)
>******* No Good Deed Ever Goes Unpunished (Mark Twain/Abraham Maslow) ******
>
>

-- Bruno Di Stefano -- Private: au843@torfree.net IEEE:b.distefano@ieee.org
Courses: stefano@ecf.toronto.edu http://www.ecf.toronto.edu/~stefano
Research: bruno@ecf.toronto.edu http://www.ecf.toronto.edu/~bruno
Consulting: alawnicz@sympatico.ca http://www3.sympatico.ca/alawnicz/nuptek.htm
-----------------------------------------------------------------------------

############################################################################
This message was posted through the fuzzy mailing list.
(1) To subscribe to this mailing list, send a message body of
"SUB FUZZY-MAIL myFirstName mySurname" to listproc@dbai.tuwien.ac.at
(2) To unsubscribe from this mailing list, send a message body of
"UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
to listproc@dbai.tuwien.ac.at
(3) To reach the human who maintains the list, send mail to
fuzzy-owner@dbai.tuwien.ac.at
(4) WWW access and other information on Fuzzy Sets and Logic see
http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
(5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html