State of Handwriting Recognition?
I need to convert old journals with messy cursive to text. I’ll be wanting to do this into the future. I get OK results with Google Vision, but was wondering if there was a decisive winner in this space?
If not, is it possible to train a particular solution to better understand my handwriting over time?
I run a document parsing service, focused on structure detection like tables and form-data. We have tested 27 OCR engines till date, and we do it every 6 months. In our latest evaluation hyperscience.com turned out to be the best in HW recognition, expensive too. Azure's cognitive service is next to it.
P.S: I've no affiliation with the mentioned services
I had good luck with Evernote, although I remember exporting to be a bit of a pain. Microsoft pioneered a few techniques over the years but I don't know if they're still near the top of the heap. The Office Lens app and OneNote should at least be competent.
Understanding handwriting at the same level as a human requires a general AI, as handwriting is usually so sloppy that it is impossible to tell what is actually written and you have to use context to decipher what it probably said.