Re: Summarization

Carlos Gershenson (carlos@jlagunez.iquimica.unam.mx)
Sun, 1 Nov 1998 22:08:26 +0100 (MET)

To try to do summarization, we should first analyse how do we summarize.

We read A. Then, we reason about it. We could say that we had to
"Translate" what we read to a "mental language". Then, we simplify it in
this mental language, and then translate it again to spoken language.

We can see that it is very hard that a system could do this. Like Prof.
Zadeh wrote, " ..."summarizing" is less vague and more circumscribed than
"learning" and "understanding." It is for this reason that I suggested
summarization as an acid test of knowledge and understanding." So, if we
agree with his viewpoint, the summarizing system would have to "know" and
"understand" before it summarizes. And artificial knowledge and
understanding are, as we say around here, in diapers. Therefore, I also
doublt of the capability of actual summarizing systems.

Regards, Carlos

On Mon, 26 Oct 1998, Michelle T. Lin wrote:

>
> To: BISC Group
> From: L. A. Zadeh <zadeh@cs.berkeley.edu>
>
>
> For your information, following is my response to some of
> the comments made about summarization.
>
>
> Warm regards,
>
> Lotfi Zadeh
>
> --------------
>
> Following is my response to the comments made by Jaime
> Carbonell, Pat Hayes, Hector Levesque, John McCarthy, Herb Simon and
> John Sowa. My response is focused on the problem of summarization.
>
> The concepts of "language learning," "understanding" and
> "summarizing" -- along with many other concepts in the domain of human
> cognition -- share a basic characteristic: they are not amenable to
> definition. More specifically, such concepts may be said to be
> amorphic in the following sense. Given a concept, C, I may be able to
> point to or generate examples of C. Furthermore, given an object, b,
> I may be able to recognize whether or not b fits my perception of C or
> the degree to which it does so. And yet, I may be unable to
> articulate the generation and/or recognition algorithms which I employ
> on a subconscious level to explain what I mean by C. This is what a
> Supreme Court justice had in mind when he said "I cannot define
> pornography but I know it when I see it."
>
> The fact that "learning," "understanding" and "summarizing"
> are amorphic concepts does not mean that they are equally vague, that
> is, equally lacking in specificity. In my view, "summarizing" is less
> vague and more circumscribed than "learning" and "understanding." It
> is for this reason that I suggested summarization as an acid test of
> knowledge and understanding.
>
> For AI, automation of summarization presents a real challenge
> -- much more so than machine translation. Among the obvious
> difficulties -- which are cited in my earlier message -- are the
> following.
>
> 1. We do not know how to define what is meant by a summary
> nor can we articulate in general terms how to assess the
> goodness of a summary.
> 2. A summary is both end-use and length dependent, and,
> ipso facto, not unique.
> 3. Summarization cannot be achieved by chunking or, more
> generally, by divide and conquer techniques.
> 4. Summarization is highly context-dependent and requires
> both local and global understanding.
> 5. Summarization by omission is not acceptable because it
> trivializes the problem.
>
> It is not a matter of dispute that -- as pointed out by
> discussants -- summarization is an active area of research and there
> exist many summarization programs. How powerful are they? Would they
> be able to come up with an acceptable approximation to the eight-word
> headline which summarized the story cited in my earlier message. For
> convenience, the story is reproduced below.
>
> A rocket exploded just yards from a convoy carrying Cambodian
> leader Hun Sen today, killing four people - including two children -
> but leaving Hun Sen uninjured. An aide called it an assassination
> attempt. Three other people were injured by the explosion, which
> occurred as Hun Sen was en route to the king's residence in the
> north-western town of Siem Reap for the beginning of ceremonies to
> swear in a new parliament. The swearing-in proceeded as scheduled.
> Hun Sen looked unperturbed after arriving at the residence of King
> Norodon Sihanouk, who spoke to the 122 lawmakers before they headed to
> the famous 12th century temple Angkor Wat for the ceremonies. Colonel
> Than Chay, local deputy police commander, said the lead vehicles of
> Hun Sen's security detail had already passed by and the car carrying
> Hun Sen was less than 10 yards away when the explosion occurred.
> Police said three more unexploded rockets were found at the side of
> the road.
>
> In this spirit, I should like to suggest the following simple
> test of summarization capability of a program, p.
>
> Select a story, s, at random from an issue of a newspaper.
> Apply p to s, yielding a summary S(s), which is of the same length as
> the headline, H(s). Compare S(s) with H(s) to assess the goodness of
> S(s). Would the available summarization programs do well on this
> test?
>
> Incidentally, in the case of the Turing test, the ability or
> inability to summarize would reveal very quickly whether on the other
> side of the curtain there is a human or a machine.
>
> Needless to say, this is much more to summarization than
> summarization of stories, articles and books. Summarization -- as a
> human form of data compression -- plays a crucial role in learning,
> understanding and acquisition of knowledge and experience. A better
> understanding of summarization is essential to enhance our ability to
> conceive, design and build machines that are much more intelligent
> than those we have today.
>
> Warm regards to all,
>
> Lotfi
>
> ------------------------------------------------------
> Lotfi A. Zadeh
> Professor in the Graduate School and Director,
> Berkeley Initiative in Soft Computing (BISC)
> CS Division, Department of EECS
> University of California
> Berkeley, CA 94720-1776
> Tel/office: (510) 642-4959 Fax/office: (510) 642-1712
> Tel/home: (510) 526-2569 Fax/home: (510) 526-2433
> email: zadeh@cs.berkeley.edu
> ------------------------------------------------------
>
>
> ############################################################################
> 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
>

"About a fact there are as many different truths
as there are different sets in the universe...
and lies, much more."

Carlos Gershenson
http://132.248.11.4/~carlos/

############################################################################
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