Although I totally agree that people should set their scanning software at settings that seem to achieve the best results, which requires the scanner to actually look at their results, it is not possible to completely illiminate those types of problems. Your points are good problems to address, but the solutions aren't that simple in all cases. I think the hyphen problem is the most solvable with good software, but obviously not everyone can afforde it. The split and joined characters can occur in a book even when settings are good, but there should not be large numbers of these mistakes. It is just that with some books when you turn brightness up you get more of one kind of mistake and when you turn it down you get another kind of mistake. Then the only thing you can do is decide which is easier to fix, and in some cases turn to using grey scale if your system can handle it. I have no idea how to prevent large areas of junk characters caused by diagrams, but I can delete them. If you use the RTK engine in Kurzweil you are much less likely to get these messes because it doesn't seem to try as hard as fine engine to make everything in to text. Of course this causes text to be given up on when it shouldn't be some times, so everything is basically a compromise, and a scanner must learn to judge what is probably the best for each scan. I haven't noticed that optimizing always solves the problems either. I have never found a scan without mistakes in interpreting 1 7 l | / and the vertically oriented like. And by the way, scanners get $2.50 not $3.50. Sarah Van Oosterwijck curious entity at earthlink dot net ----- Original Message ----- From: "E." <thoth93@xxxxxxxxxxxxx> To: <bksvol-discuss@xxxxxxxxxxxxx> Sent: Saturday, August 14, 2004 2:13 AM Subject: [bksvol-discuss] Re: Requirements for acceptance -- the bottom line > I almost totally do validating rather than scanning. perhaps what I ask > below is asking a lot. I am, however, concerned about the following > patterns which occur a lot. Perhaps someone can talk about how to set a > scanner so they appear less often. The reason I keep harping on folks > getting better scanning software (that is the new upgrades) is that some of > these scanning settings can then be set by the scanning software > illiminating these patterns. > > rn for m as in bam for bar, cornputer for computer, cornmande for command, > n or bum for burn > m for rn resulting in comer for corner > space hypens in the middle of words particularly names such as Eliz -abeth > or Eliz -abeth > large chunks of garbage text where diagrams should be which sometimes > causes the braillenote to crash during validating > I for 1 resulting in i8i5 for 1815 > 1 for I resulting in 1 am instead of I am > > A validator get gets rid of these using find and replace judiciously of > course. It still takes a lot longer to do this than it does to get the > scan right in the first place. > > Finally remember those who scan get about three fifty worth of credit per > scan. Validators get fifty cents per validation. Not a big point but > certainly one which demonstrates the dedication of validators. > > Even use of rank spelling could mean some of these errors can be fixed > before the book is uploaded to the step 1 page. If I can use rank spelling > to fix cornmand into command so can a volunteer scanning a book. > > E. > > >