Proofreading, part 21

Even this process, though, isn't completely foolproof. Humans run the software, and as soon as humans come on the scene, there's potential for human error.

If the various software operations aren't applied properly - like forgetting to run the spellcheck on a document - the software won't flag up problems because it won't have looked for them. Perhaps the “is it there?" process on each letter of a column was run but the “is it kosher?" process accidentally wasn't.

The computer needs human help to learn the writing, and perhaps the human isn't paying attention. Perhaps the computer says “hey, human, what's this?" and the human is half-asleep and says “vav" when he means “yud," and a spelling mistake consequently goes unspotted.

Sometimes the software just can't pick up on things. Very fine lines - the scanner might not pick them up; sometimes the presence or absence of a very fine line can be the difference between kosher and pasul. But we can't (at present) scan to so high a resolution as to pick up on all these; the processing time would be prohibitive.

Finally, the letters are very slightly three-dimensional; a human, with stereo vision, can tell the difference between ink and shadow, and a scanner can't always. Sometimes it'll interpret a shadow as a crucial fine line, and report a letter kosher when it really isn't.

So a scan is an excellent tool - I think it's one of the finer syntheses of technological development and ancient ritual - but it does not replace all the other proofreading tools we use, and it is not a substitute for hard work and knowing your stuff. Few things are, really.