[bksvol-discuss] Re: Quality checks procedure -- was WRe: Re: Self-validation
- From: "Gary Wunder" <gwunder@xxxxxxxxxxxxx>
- To: <bksvol-discuss@xxxxxxxxxxxxx>
- Date: Fri, 18 Jun 2004 14:02:53 -0500
Nice list - thank you.
----- Original Message -----
From: Guido Corona
To: bksvol-discuss@xxxxxxxxxxxxx
Sent: Friday, June 18, 2004 1:55 PM
Subject: [bksvol-discuss] Quality checks procedure -- was WRe: Re:
Self-validation
I will not get into the self-validating debate. But I suggest that both
scanners and validators go through a set of quality checks that are as
mechanized as possible, to avoid the 'my book' vs. 'someone else's' book
problem. I found that the following set of checks tend to generate rather high
results:
1. Page Integrity Check: sample every 20 pages.
2. Chapter header check: works best if the book uses the word 'chapter' or
something else to search for.
3. Page header check: lately I have been stripping headers manually and
leaving only page numbers. Definitely tedious, as I do it on each and every
page.
4. during the previous step I also check for the first word of each page.
If the word looks like a fragment, I merge it with the last word on the
previous page if appropriate.
5. Search for short dash followed by space, and short dash followed by
newline. These will let you find all sorts of words that were split at end of
lines or at end of pages and can be repaired.
6. Search for the tab char (\t in k1k). This is most often a junk char,
especially abundant at the beginning and end of lines. You will frequently
find it associated with other junk chars, or single alphabets that had no
business being there. Remove manually each occurrence of these clustered nasty
things as appropriate.
7. Junk char hunt. Look for junk chars available from the keyboard. Start
from the top left of the keyboard and work your way down to the bottom right.
Remove or repair manually as required.
Jim Pardee also suggested we keep a file containing those chars that are not
keyboardable: we can copy/paste them in the find dialogue to search for them
in the document.
9. Look for whole words consisting of digit '1'. In many cases you should
change them to 'I'. Sometimes they are to be deleted. Do each change manually
as appropriate.
10. look for digit 1 followed by an apostrophes: in most cases that should
be changed to I followed by apostrophes.
11. look for the apostrophes followed by digit 1. In most cases that is
part of a '11, which should become an 'll.
12. Do a mass replacement of double single quote with single double quote.
13. Remove/fix single alphabetic words: start with 'b'. Search should be
capitalization insensitive, except for 'i' which should be searched in lower
case only. Delete or repair each occurrence manually as appropriate. Be
careful, you may be deleting someone's middle initial.
14. Spell check: This step should remove most residual problems, except for
some scanos that have generated valid English words.
Hope this helps.
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
Nolan Crabb <aa3go@xxxxxxxxxxx>
Sent by: bksvol-discuss-bounce@xxxxxxxxxxxxx
06/18/2004 12:23 PM Please respond to
bksvol-discuss
To bksvol-discuss@xxxxxxxxxxxxx
cc
Subject [bksvol-discuss] Re: Self-validation
Rui warned against the urge to self validate. I completely concur!
I've made a living editing my work and that of others for years. The harsh
truth is, your own errors are much easier to miss, even if you've let that
book sit up there and cool its digital heels for days. I guess the urge to
self validate is a natural one, since people get submission credits and
such that will help them pay for next year's subscription. I have a Kate
Wilhelm mystery that's been up there for some time now, and I want very
much to just validate the thing and get the credits and more importantly,
get it up there so others can enjoy it. But I won't. I'm too aware of how
easy it is to skip errors in things you've either written or read. I think
the checks and balances that exist here--the ones that encourage others to
validate what you've submitted--are the way it should work. I realize
others will challenge my position, suggesting that self validation is
absolutely the only way some of the more esoteric titles will get
approved. I disagree. The first book I ever validated was a Christian
romance--decidedly not, not, not something I would normally want to read
under any circumstances. Oddly enough, that's precisely the reason I chose
it. I figured the material would be so new and different to me that I'd be
more prone to catch errors. That book entered the Bookshare system with a
"good" rating presumably provided by the submitter. I spent some time with
the book, but today it carries an "excellent" rating, and it's now part of
the collection.
Please try not to misinterpret this, folks. I don't use it as an example
to demonstrate how amazing I am. Very nearly all of you have been at the
submission and validation end of this far longer than have I, and you're
doubtless the ultimate experts, having forgotten more in a day than I will
learn in years. I just find self validation a little scary, especially in
light of rather strong messages lately which have called for higher quality
scans and validations. There's no doubt we achieve higher quality
validations if we don't do them ourselves.
The quarterly magazine I edit goes through no fewer than four different
edits before it ever sees the inside of our subscribers' mailboxes. I'm
not advocating for absolute rigid perfection; we are volunteers, after all,
who have lives. But self validation is an excellent way to increase the
number of potential errors into the system.
So that I don't totally come across here as being the loud mouthed whiner
on the list, here's a little proposal: If you have a book that's been up
there quite a while, I'll take yours and validate it, regardless of the
subject or whatever, if you take mine and get it approved. It's called
"The Casebook of Constance and Charlie Vol. 1," and it's 614 pages, so I'm
sure that's discouraged more than one person from taking it. Obviously,
this is one of those first-come first-accepted challenges. <smile>
Again, I'm not desirous of offending any here. But in light of recent
messages that have called for higher standards in terms of better quality
scans and better validations, redoubling our resolve to let others validate
our work is probably one good way to ensure the increased quality of the
collection.
Best Regards,
Nolan, who is dawning his fire-retardant e-mail-reading suit in preparation
for all that indignant mail from self validators :-)
Other related posts:
