How to understand Kobo highlights

[geekraver]

Hi all

I just finished reading a friend's manuscript for his novel. He gave it to me as an epub on request, and I read it on my Kobo.

While reading I found many typos and grammatical errors. Whenever I found one I would highlight a word at or near the problem. I'm now realizing that that was a bad idea as the exported highlights are mostly meaningless (I have many "the" and "and" etc type words highlighted). I could go through them all and expand the highlight but it would be very tedious as there are close to 200.

I thought that perhaps I could figure out how to get more context by looking at the raw data in the KoboReader.sqllite file. What I see there is a bunch of items of a form like this:

untitled.html#point(/1/2/1/6833/1/1:431)

My question is: does anyone know how to interpret that? If I extract the untitled.html from the epub, how can I use that to get the surrounding context?

Thanks!
Graham

[cramoisi]

I won't be of much help but when opening your thread mobileread propose me this thread is the "similar threads" section. I skimmed it and it seems it would be of some help.

https://www.mobileread.com/forums/sh...d.php?t=258753


(for my experience, it's not a good idea to readproof on epub (even if I would prefer to do it in a ideal world) because after that it's difficult for the author to find the sentences / words you highlighted without being able to match them on a page number in their word-processor)

[davidfor]

Firstly, I do lots of highlighting for a similar purpose. I found the best way to do this was to highlight a few words around the problem word. I usually make sure the problem word is the first. And I might add an annotation to remind me of exactly what the problem is.

For location stored in the bookmarks table, it is a walk of the tags in the file the highlight is in. Each slash represents a tag level inside the file. For your example:

untitled.html#point(/1/2/1/6833/1/1:431)

The file first part is the file name in the epub.

The numbers work by counting tags. The first "1" is the "html". The "2" is the body ("head" then "body" within "html"). The next one probably means there is a div or something enclosing all the text. The 6833 is the count of tags within this. But, it is actually the count of opening and closing tags at this level. It means the highlight is within the 3412th open tag at this level, counting all headings, paragraphs, rules etc. Then there is another open tag of some sort, probably a span or some sort of emphasis tag. I'm not sure what 1 before the colon is, but the last number is where the highlight starts within the tag: the 431st character.

I might have one of the early levels wrong. It's been a while since I did this and I seem to remember something inconsistent in how one of these numbers worked. But the big number is correct and effectively means the 3412th paragraph but includes things that you might not normally consider paragraphs like headings and rules.

Not that's a pain and probably not worth doing anything with. I've done this manually to test how this works, but stuck to the beginning of the file. I kept losing count.

But, that's not what you want to do. What might work for you is Toxaris's Kobo annotation exporter. It has been a while since I last used it, but my memory is that it extracts the surrounding text when getting the annotations and highlights that in the file it produces. I don't know if it still works with recent Kobo firmware. That will depend on what SQLite library it uses.

There are a number of other ways to get the annotations from the Kobo devices. I can't remember if any of the others will get the context. The calibre Annotations plugin definitely doesn't, nor the export that can be enabled on the device.

[RobertJSawyer]

Another alternative is to capture screenshots of your highlights (I actually do this a lot for research, and dump the screenshots into Evernote, which OCRs them and makes them searchable).

This message explains how to enable screenshots:

https://www.mobileread.com/forums/sh...&postcount=444

Then just give your friend all the screenshots, and he can review them with IrfanView or any other graphics viewer.

[geekraver]

Thanks. As a Mac owner I can't use the VB program unfortunately. I've mucked about in Python to and managed to extract the paragraphs I needed.

[geek1011]

Also, if you do this again, you may want to convert your books to kepub. The locations stored in the database are a lot more readable (basically file.html#kobo.paragraph.sentence, and way easier to extract (they use IDs).