OCR in reader?

[wibr]

So I am scanning some Chinese books and obviously I would like to read them in Pleco, using the integrated OCR.

I know everything is pretty busy right now, so this is just my wishlist.


Load pdf/zip/some containerfile with all the images as pages

Set zoom level and lock to vertical scrolling

Save two scrolling positions. This is for those exercise books where the solution is somewhere at the end of the book/chapter. Or textbooks with vocabs at the end of the chapter/book. I imagine it to work like this: There is a small button somewhere toggling the scrolling positions, e.g. first I am doing the exercises in state 1, then I want to look up the solution, so I switch to state 2, when I do this the first time I have to scroll all the way down to the solutions, but going back to the exercises I just have to toggle to state 1 again and the scrolling position for the exercises is restored...

Tap on a character to OCR it and then show the info bubble, just like in normal reader. Tap with two finger to OCR everything in between and show a bubble if it's a word in one of the dictionaries

 

[mikelove]

Load pdf/zip/some containerfile with all the images as pages

Unfortunately, we haven't yet found a PDF library we can license for Android that supports Chinese consistently and doesn't come with an outlandishly high royalty. (we are supporting PDF in OCR on iOS in our next update, but PDF support is free and built-in on iOS)

Set zoom level and lock to vertical scrolling

Have you tried tapping on the button to lock the recognition area in place? You can do that, zoom it to the whole document and then just scroll around it tapping on words to look them up.

Save two scrolling positions. This is for those exercise books where the solution is somewhere at the end of the book/chapter. Or textbooks with vocabs at the end of the chapter/book. I imagine it to work like this: There is a small button somewhere toggling the scrolling positions, e.g. first I am doing the exercises in state 1, then I want to look up the solution, so I switch to state 2, when I do this the first time I have to scroll all the way down to the solutions, but going back to the exercises I just have to toggle to state 1 again and the scrolling position for the exercises is restored...

Doable, but the UI is tricky - extra buttons / overlays / etc - so I'd be inclined to wait until after we've increased our hardware spec / OS requirements a bit and can therefore worry less about the consequences of those changes on a potentially-laggy / low-memory OCR session.

Tap on a character to OCR it and then show the info bubble, just like in normal reader. Tap with two finger to OCR everything in between and show a bubble if it's a word in one of the dictionaries

That one's actually similar to something a few iOS users have requested too.

 

[wibr]

Hi Mike,

yeah you already mentioned the difference between pdf on Android and iOS, that's why I think a zip file would also do it. If there is not enough demand for it I can see that's very specific (only relevant for scanned books) and not worth implementing, especially since there will be pdf support on iOS...

The basic idea was to make the reading experience (especially for scanned pages) more convenient, using the OCR loading each page separately and then moving the page around the fixed OCR field is a bit cumbersome...

Well as always I am looking forward to future versions, mostly Android but so far I've done all the purchases on my iPod, just in case I might end up with an iPad someday...

 

[mikelove]

yeah you already mentioned the difference between pdf on Android and iOS, that's why I think a zip file would also do it. If there is not enough demand for it I can see that's very specific (only relevant for scanned books) and not worth implementing, especially since there will be pdf support on iOS...

A directory would be easier than a .zip, actually, but in either case you've got the problem that there's no clear sequence or structure to book pages - we'd either need to enforce a rigid file numbering system or support some kind of XML "manifest" that hardly anybody would actually go to the trouble of coding. And honestly at that point I'd rather focus on EPUB, particularly since it seems likely that the vast majority of textbooks will be available in that format within the next few years - for our big iOS update we've got a new way of efficiently looking up text "live" on web pages (which is what EPUBs actually are, a series of HTML files), and unlike our old way of doing it we think think this method can also work on Android.

The basic idea was to make the reading experience (especially for scanned pages) more convenient, using the OCR loading each page separately and then moving the page around the fixed OCR field is a bit cumbersome...

Yes, that certainly makes sense - if PDF OCR support on iOS proves to be a success that'll make this a good bit easier to justify.

 

[stephanhodges]

Don't forget that it's relatively easy to convert MOST pdf files to epub using the free software 'calibre'.

 

[mikelove]

Don't forget that it's relatively easy to convert MOST pdf files to epub using the free software 'calibre'.

True, and actually in our next iOS update you won't even need to OCR those PDFs to use them in Pleco - they'll support tap lookups just like other files do. (we can probably do that on Android too if we could just find somebody with a good PDF reader library whose licensing demands are within what we can reasonably afford)

 

[wibr]

in our next iOS update you won't even need to OCR those PDFs to use them in Pleco - they'll support tap lookups just like other files do.

Not sure if I understand that correctly, are "those PDFs" normal pdfs with text as text or text as pixels (using the Pleco OCR for recognition)?
If you use OCR in the reader anyway, couldn't you offer the same functionality for images which are embedded in epub?

 

[mikelove]

Not sure if I understand that correctly, are "those PDFs" normal pdfs with text as text or text as pixels (using the Pleco OCR for recognition)?

PDFs with text, the same ones that can be converted to EPUBs / TXT files with Calibre.

If you use OCR in the reader anyway, couldn't you offer the same functionality for images which are embedded in epub?

Yes, but extracting the actual embedded text from the PDF is much more reliable / easy.