finding l's OCRed as I's

[jlocicero]

Using Calibre's edit book function, I would like to search for any letter l (ell) that has been incorrectly OCRed as an I (capital i), and then replace it with an l. Any easy way to do this?

I thought if I could search for "(any letter) + I", I would be able to weed out any legitimate uses of "I". Is this a good method? How would I do this in Calibre?

Thanks!

[eschwartz]

use regex mode (and lookbehinds, which do NOT capture):

(?<=[a-z])I

Replace with "l"
This won't catch the illegitimate word "Iegitimate", though.

Hmm,

(?<![."]\s*)I

That should match any "I" that does NOT follow punctuation (even if there is an intervening space).

I'd still eyeball each edit first -- or scan each change using "See what's changed".

[DomesticExtremis]

Assuming it is english (where only proper nouns are capitalised), then eschwartz's suggestion will catch many.

Also, for words starting with ell:

PHP Code:

search: I[a-z]
replace:l\1 


It means two searches

[eschwartz]

Originally Posted by DomesticExtremis

Assuming it is english (where only proper nouns are capitalised), then eschwartz's suggestion will catch many.

Also, for words starting with ell:

PHP Code:

search: I[a-z]
replace:l\1 


It means two searches

I have updated my regex to match "I" when not following punctuation, which is better than doing two searches that together equal

I

[DomesticExtremis]

It might just be easier to switch to a serif font (for the purposes
of seeing the wrong 'uns) and just running the spellchecker.

[Tex2002ans]

Originally Posted by jlocicero

I thought if I could search for "(any letter) + I", I would be able to weed out any legitimate uses of "I". Is this a good method? How would I do this in Calibre?

I think a much more efficient way to do it would just be to take advantage of the "Spell Check" functionality. (I use this ALL THE TIME in Sigil).

I mean, it already lists every single word in the book, why not take advantage of that?

Just type a capital letter 'I' in the search, and you should be able to see every single word with the letter 'I' in it:

Here is the Calibre Manual on the Spell Check functionality:

http://manual.calibre-ebook.com/edit...ds-in-the-book

Side Note: The Spell Check tool is also extremely helpful to catch hyphens + accented words + lots of other stuff:

https://www.mobileread.com/forums/sho...84&postcount=4
https://www.mobileread.com/forums/sho...08&postcount=7

Potential Calibre Enhancement: It would be nice to also be able to toggle a Case Sensitive SEARCH. This could definitely make the Spell Check tool much more useful than Sigil's implementation. (I daresay, this might just be useful enough to make me jump ship.... maybe. )

For example, here is a book in Sigil showing off all of the misspelled (not in a dictionary) words with a lowercase 'i' and an uppercase 'I':



It would be nice to see the words with ONLY the uppercase version. (For example, finding "Ice-Cream" but not "ice-cream")

[theducks]

Originally Posted by Tex2002ans

I think a much more efficient way to do it would just be to take advantage of the "Spell Check" functionality. (I use this ALL THE TIME in Sigil).

I mean, it already lists every single word in the book, why not take advantage of that?

Just type a capital letter 'I' in the search, and you should be able to see every single word with the letter 'I' in it:

Here is the Calibre Manual on the Spell Check functionality:

http://manual.calibre-ebook.com/edit...ds-in-the-book

Side Note: The Spell Check tool is also extremely helpful to catch hyphens + accented words + lots of other stuff:

https://www.mobileread.com/forums/sho...84&postcount=4
https://www.mobileread.com/forums/sho...08&postcount=7

Potential Calibre Enhancement: It would be nice to also be able to toggle a Case Sensitive SEARCH. This could definitely make the Spell Check tool much more useful than Sigil's implementation. (I daresay, this might just be useful enough to make me jump ship.... maybe. )

For example, here is a book in Sigil showing off all of the misspelled (not in a dictionary) words with a lowercase 'i' and an uppercase 'I':

Attachment 124343

It would be nice to see the words with ONLY the uppercase version. (For example, finding "Ice-Cream" but not "ice-cream")

I just tick the case sensitive sort (Sigil and Calibre)
A lower case word with a capital in the middle sticks out like my thumb after a misaimed hammer blow

[jlocicero]

Originally Posted by eschwartz

use regex mode (and lookbehinds, which do NOT capture):

(?<=[a-z])I

Replace with "l"
This won't catch the illegitimate word "Iegitimate", though.

Hmm,

(?<![."]\s*)I

That should match any "I" that does NOT follow punctuation (even if there is an intervening space).

I'd still eyeball each edit first -- or scan each change using "See what's changed".

I just tried both of these, and they find all i's in the text. I made sure Regex mode was selected. What have I done wrong?

[eschwartz]

Originally Posted by jlocicero

I just tried both of these, and they find all i's in the text. I made sure Regex mode was selected. What have I done wrong?

I am guessing "case sensitive" is not checked.

By default it seems to be off.

[Tex2002ans]

Originally Posted by theducks

I just tick the case sensitive sort (Sigil and Calibre)
A lower case word with a capital in the middle sticks out like my thumb after a misaimed hammer blow

Just visualizing the words in a different way helps you spot errors! (sort ascending, sort descending, sort by frequency, sort by case, etc. etc.)

I typically do at least two passes when using that tool during my error checking phase. One with case sensitive sort checked, and one without.

Also, you can sort by frequency. Typically an OCR error only occurs a handful of times (1-3), so you can rule out many of the more common words.

[mrmikel]

search: I[a-z]
replace:l\1

I have found for this to work in calibre editor it has to be search: I([a-z])
in order to pick up the following letter.

AND even though you are indicating small letters [a-z], it will pickup capital and lowercase unless case sensitive is also checked. In this way it is not exactly as you would expect.

[eschwartz]

Originally Posted by mrmikel

search: I[a-z]
replace:l\1

I have found for this to work in calibre editor it has to be search: I([a-z])
in order to pick up the following letter.

AND even though you are indicating small letters [a-z], it will pickup capital and lowercase unless case sensitive is also checked. In this way it is not exactly as you would expect.

That will still pick up every single instance of a capital "I" in the book, as long as it isn't the actual word "I".

If you leave off the captured "([a-z])" then you would get the same results (except for the word "I"), and could replace with "l".

Much better to look for anything-but-punctuation before the "I"... as I did in post #2.

And negative lookbehind means you don't have to capture anything, you can just replace the letter itself.

[Jellby]

An initial L is probably followed by a vowel, an initial I is probably followed by a consonant. Try searching for "I[aeiouy]".

[mrmikel]

That is a very good suggestion. Io is probably the only one on which it would pick up something unnecessary and unless the book is about chemistry or space, it should not be a big deal.

[jlocicero]

Originally Posted by eschwartz

I am guessing "case sensitive" is not checked.

By default it seems to be off.

Yes, my mistake, and now it works.

Thanks for all the suggestion! I've gotten better results than I hoped for.