[bksvol-discuss] Re: Miraculous multiplication of book pages -- a bug of biblical proportion?
- From: "Kenneth A. Cross" <crossk@xxxxxxxxxxxx>
- To: <bksvol-discuss@xxxxxxxxxxxxx>
- Date: Tue, 10 Aug 2004 18:26:52 -0400
No, I don't have 3000. I sent Stephen the entire file, and we are talking
about it.
----- Original Message -----
From: Guido Corona
To: bksvol-discuss@xxxxxxxxxxxxx
Sent: Tuesday, August 10, 2004 2:45 PM
Subject: [bksvol-discuss] Re: Miraculous multiplication of book pages -- a
bug of biblical proportion?
Thanks Ken. You do not happen to be using K3000 rather than K1000 by any
chance?
Let Stephen know exactly which product and version you are using.
It may also be a good idea to send Stephen a few sample images of pages that
give the most bizarre results. E.g. title pages where it seems each char in
the title is assigned its unique page.
Thanks,
Guido
Guido D. Corona
IBM Accessibility Center, Austin Tx.
IBM Research,
Phone: (512) 838-9735
Email: guidoc@xxxxxxxxxxx
Visit my weekly Accessibility WebLog at:
http://www-3.ibm.com/able/weblog/corona_weblog.html
"Kenneth A. Cross" <crossk@xxxxxxxxxxxx>
Sent by: bksvol-discuss-bounce@xxxxxxxxxxxxx
08/10/2004 12:59 PM Please respond to
bksvol-discuss
To <bksvol-discuss@xxxxxxxxxxxxx>
cc
Subject [bksvol-discuss] Re: Miraculous multiplication of book
pages -- a bug of biblical proportion?
I have sent them the kes file to study. The kes file of about two hundred
pages converted to 500 plus in rtf. The same kes file converted to the same
two hundred pages when I changed the kes file to txt. I will also get the
settings file to them.
----- Original Message -----
From: Guido Corona
To: bksvol-discuss@xxxxxxxxxxxxx
Sent: Tuesday, August 10, 2004 12:38 PM
Subject: [bksvol-discuss] Miraculous multiplication of book pages -- a bug of
biblical proportion?
Ken, that worries me. I had not heard yet of K1000 Kes to RTF conversion
fowling page boundaries this badly . I am contacting Kurzweil on this
subject. They might need the settings file (OST) you used to reproduce the
problem.
The only setting which in my view may be the cause of these extra pages is in
recognition settings: partial columns should be ignored rather than kept.
In the meantime, are you saying that a further K1000 conversion from RTF to
TXT cleans up the spurious page breaks? If that is the case, I will try this
in the next book where I encounter the problem.
I suggest that until we get the problem fixed programmatically, in the
meantime you convert problem books to TXT before submission to Bookshare.
Guido
forassistance. If we can provide with reproduceable scenarios, I am sure
they can address the problem
Guido D. Corona
IBM Accessibility Center, Austin Tx.
IBM Research,
Phone: (512) 838-9735
Email: guidoc@xxxxxxxxxxx
Visit my weekly Accessibility WebLog at:
http://www-3.ibm.com/able/weblog/corona_weblog.html
"Kenneth A. Cross" <crossk@xxxxxxxxxxxx>
Sent by: bksvol-discuss-bounce@xxxxxxxxxxxxx
08/10/2004 11:01 AM
Please respond to
bksvol-discuss
To <bksvol-discuss@xxxxxxxxxxxxx>
cc
Subject [bksvol-discuss] Re: 550 books in the download queue
Actually, there is a very interesting problem, at least in Kurzweil 8.1. I
just saved a 219 page book from kes to rtf. The rtf version had five hundred
plus pages. More pointedly, after deleting the pages and saving the book, the
pages reappeared when the file was loaded and saved again. That had nothing
to do with the scanning and very little to do with the actual reading of the
book. Now that same book saved as a two hundred page txt book, and if you want
books in that format that is easy enough to do.
----- Original Message -----
From: Guido Corona
To: bksvol-discuss@xxxxxxxxxxxxx
Sent: Tuesday, August 10, 2004 11:34 AM
Subject: [bksvol-discuss] Re: 550 books in the download queue
Kenneth, I do agree that we cannot make things perfect, especially if
graphs and pictures are involved, which typically generate optical noise. On
the other hand I do urge all scanning volunteers to perform a modicum of
cleanup of their materials prior to posting to the system.
1. PLEASE do perform a book integrity check. If a print copy has 320 pages,
there is no reason whatsoever that the etext should contain 1760 pages instead.
There should be a 1 to 1 correspondence in the bodytext between printed pages
and etext pages. Any volunteer can in most cases easily remove some duplicate
pages, but can't fix things when they are totally out of kilter. Furthermore,
it is a lot easier for the submitter to integrate any missing pages as he/she
has still access to the print copy.
In most cases, even if a few missing pages need to be inserted, a page
integrity check for a book does not take more than 15 minutes and would save
all reviewers a lot of wasted time.
In some cases there is -- as already mentioned -- a tremendous discrepancy
between pages in the printed book and etext pages, where the etext has 5 or 6
times the number of pages in the original. This likely points to scanning/OCR
settings that are way off, or an OCR package being used which is less than
adequate.
2. Lots of broken page headers make a book very tiring to read. Please fix
them or remove them. Kurzweil lets you remove page headers automatically.
Version 8 was a little radical in this regard and ended removing also stuff it
was not supposed to. Newest version 9, just announced yesterday, has now an
option for 'careful' header removal. Yesterday I worked on one of your books.
Kurzweil removed 190 headers. Approx 120 headers I removed manually. It took
me all of 15 minutes to do the cleanup. As I had already performed a page
integrity check and had come up with perfect correspondence after removal of
duplicate pages, I also removed page numbers to do a faster job. Just want to
get the backlog down quickly.
3. Ah yes, those amazing synopses. Let us all try to be informative . The
short one should give our paying customers a very brief sense of the book. If
we are inclined to give more detailed info, a longer and useful description
can go in the long synopsis. I confess I hardly ever make up my own synopsis,
but I liberally borrow from the front matter of the book, the back cover of
paperbacks and the front and back flaps of hard covers. Synopsis such as "Set
in Alabama", "It's all in the Title", "Thorough Treatment of the matter at
hand" are unfortunately not noticeably helpful and will only cause our paying
customers to get irritated and lose faith in Bookshare.
Of course, there is a lot more that submitters can do to improve the quality
of their postings, but even a little of upfront cleanup and a clean submission
process will enable us to offer a quality product to OUR PAYING CUSTOMERS.
NOTE: By the way, Kurzweil will start shipping K1000 version 9 shortly. I
have used the beta and found OCR quality even improved over earlier versions.
I will post the announcement shortly. Cost of the upgrade is $95 or $0.00,
depending on the status of your account.
Guido D. Corona
IBM Accessibility Center, Austin Tx.
IBM Research,
Phone: (512) 838-9735
Email: guidoc@xxxxxxxxxxx
Visit my weekly Accessibility WebLog at:
http://www-3.ibm.com/able/weblog/corona_weblog.html
Other related posts:
