I use Devonthink primarily for a database that consists of photos of some ancient handwritten diaries. The handwriting is usually neat and an Italian. Is there anyway to get the handwriting in the photo “recognised” and turned into Searchable PDF’s?
DEVONthink does not have that technology included. However there are numerous applications out there that say they have that capability which most likely can be operated in unison with DEVONthink. I saw a number found via Internet search “handwriting to pdf”. DEVONthink’s important feature is its skill at integration with other apps.
Once in PDF, DEVONthink provides OCR capabilities.
Which ones did you see that looked promising?
I did not do any evaluation, testing, or research. My only experience with handwriting detection is with Notability and given that I can’t even discern my own handwriting, it was not an endearing experience.
1 Like
OCR of handwriting is still a holy grail, and one unlikely to ever be a high-percentage process. If you are referring to a fairly strictly drawn set of characters, it’s possible in some cases. Cursive is off the table with any package I’ve ever seen.
Thanks
THanks.I think I resemble your remarks!
Certainly not “off the table”. Pen to Print has an iOS app that will OCR cursive script from an image with pretty good results, though it does send the data online to do the magic with its API (but there again, so does DEVONthink’s clipper’s “clutter-free” thing).
There is no cursive shown in the screencaps for this app. This is not cursive…
have you actually used this and written in cursive?
Yes I have used this from written, cursive handwriting, both from the iOS camera and from an image.
From the iOS camera: in Dropbox as too big to post here.
From an image in iOS Photos app: again, in Dropbox due to file-size limitations here.
1 Like
Thanks very much. I downloaded the free version yesterday and tried it on a sample page. I was worried it would not do the job as the diary pages are in Italian and use a lot of Ethiopian place names & words and occasionally some Amharic script on some pages. It worked well on the Ipad. The sample page had mistakes but they were not as many as I expected. In fact since I have my pictures on my desktop, I then tried downloading the app on my Mac mini (M1: which can use iOS apps) and it worked (even better!) on the desktop. The Desktop app recognised words that the iPad app didn’t! This is very good… It will be interesting to see what it does with pages written more sloppily written and more confusing but I would say this was a fantastic first step. Thank you for pointing me to it.
Thanks very much. I downloaded the free version yesterday and tried it on a sample page. I was worried it would not do the job as the diary pages are in Italian and use a lot of Ethiopian place names & words and occasionally some Amharic script on some pages. It worked well on the Ipad. The sample page had mistakes but they were not as many as I expected. In fact since I have my pictures on my desktop, I then tried downloading the app on my Mac mini (M1: which can use iOS apps) and it worked (even better!) on the desktop. The Desktop app recognised words that the iPad app didn’t! This is very good… It will be interesting to see what it does with pages written more sloppily written and more confusing but I would say this was a fantastic first step. Thank you for pointing me to it.
I’m glad you found it useful. I don’t have an M1 Mac so I can’t test it the way you did, but it sounds promising. I need some new hardware! I wonder how it compares with Apple’s Scribble. Their API might be interesting.
I found TextSniper. Not sure it will do the trick for you but it works really well on texts manuscripts even on photos with background.
The idea is that you take a screenshot with cmd-shift-2 and it copy the text in the clipboard.
THanks…will look into it.
I like Notability quite a lot.
1 Like
Trouble is with Notability, as with GoodNotes, is that it only recognises handwriting that was input by the app, whereas Pen to Print takes it from external sources.
TextSniper is an extremely useful tool, but I’m still interested to tease apart Pen to Print’s API if I can.
Interesting that you were able to get it to read Italian from an ancient diary…this is what Pen to Print generated (without editing) for a clearly handwritten (US Civil War) pension application page from the late 1800s.
Test image:
Test results:
My favorite? “30. Cert of Bibobeling show soldion became infich fordelen”… for “3d. Cert of Disability shows soldier became unfit for duty.”
Still stuck with manual transcription over here. But DEVONthink’s sorter (in my menu bar) is handy for transcribing records I don’t want to download, even if no OCR program is yet capable. If I ever update my computer, I’ll definitely check out TextSniper, though. I pretty much try them all.
Interesting. Can you expound on this a little?
The sorter is so nifty being able to flip down over a website, and flip up to reposition the image/PDF, and flip down to keep going. I don’t have to split my screen and lose real estate for records that are transcribed easier when I can see the whole line/portion for context instead of just chunks. (Future feature request: allow typing comments when you take a screenshot of something.)
If I transcribe something, it first gets assigned an ID in Bookends, which I then use in every piece of information I generate for that source (both in the DT3 Name and BookendsUID metadata). If I happen to download an image/PDF later (via Bookends), I just move the text into an annotation template I use.
This method helps me avoid the Collector’s Fallacy. (DT3 is helping me sort through a decade’s worth of downloaded images that I never processed. On a related note, moving my sources into Bookends instead of relying on genealogy software was one of the best decisions ever.)
Flipping the sorter like this also works great for making lists of records I want to check out from the old index books.
(I’m playing with using formatted notes and their pretty checkboxes instead of markdown for my index checklists. Not sure I’ll stick with it, but I downloaded Evernote Legacy & reimported my Evernote notes yesterday so I can delete Evernote forever, and the checkboxes came in so nice (where last year they didn’t come in at all)…You can see with “# image 550” that I forgot I was using a formatted note, lol, but my brain still reads it as a heading, so… Then again, the sorter’s formatted note doesn’t do checkboxes–I have to insert those manually after import, and DT3 is tetchy about the checkbox label, so who knows?)
Probably more than you wanted to know, eh?
1 Like