Making the Most of OCR

[amfeather 27]

I love Evernote's OCR capabilities. Evernote's OCR allows me to (1) take notes by hand in the classroom and (2) index and search those notes later. Unfortunately, I find that the OCR is more or less successful based on how the hand-written notes are taken. On college-ruled yellow legal pads, the OCR is around 50 percent successful (many words not recognized; some words partially recognized). On clean white printer paper, the OCR is around 90 percent successful! Clearly the kind of paper you use (and how well you write) impact how well Evernote can recognize handwriting in documents.

I want to create and capture my documents to maximize the effectiveness of Evernote's OCR. Has anyone from the community identified best practices? Should I use Evernote, Scannable, or some other service for the initial intake? Should I use lined paper or blank paper? Should I use a 0.7 or 0.5 pen? I would like to reduce the margin for error as low as possible to make the most of Evernote.

Thank you!

[DTLow 5,744]

I use an external app for hand written notes; Notability on an iPad with an Apple Pencil       
It provides a convert-to-text feature so I can correct any transcription errors   
The notes are saved to Evernote in pdf format

[amfeather 27]

11 minutes ago, DTLow said:

It provides a convert-to-text feature so I can correct any transcription errors

I would love to correct errors in the OCR transcript, especially for long-term information storage! Thank you for sharing this tool.

[PinkElephant 8,795]

Unfortunately OCR done by EN in scanned documents and pictures is not that transparent. Especially the embedded OCR in EN notes is invisible.

For Scans on my iPhone I am using the app ScannerPro. It has a build in OCR capability. You can show the resulting text as an overlay to the picture of the scan. By this it is pretty easy to see whether the OCR worked correctly.

Since I found the results to be usually very good, I only check if the scanned document was difficult to OCR, like it happens with some „elegant“ fonts.

For handwriting the app GoodNotes 5 does OCR the full document as well. This happens on the fly, all the time (if enabled in the settings) and can be embedded into a pdf that is exported to other apps, like EN. There is as well a feature to convert handwriting into text, which uses the same engine as the full OCR, but only converts on demand preselected text. It is not necessary to make this conversion to get a fully OCRed handwritten document.

GoodNotes 5 has a build in search that relies on the OCRed handwriting. It is very accurate and „finds it all“. In GN 4 This was restricted to only one notebook. With GN 5 the whole stock of notebooks is searched, which is a major improvement. It is still no match to search in EN, where all sorts of documents and own notes will be searched, including pictures containing text.