The Beth Mardutho Facebook pages just announced a major development in the quest to use OCR (optical character recognition) technology for manuscripts. Calling it Handwritten-Text Recognition (HTR), they say they have it working for Syriac manuscripts. I know others have tried this for Greek for years and have run into trouble. We can hope the lessons learned here for Syriac can be passed on and used on Greek. If anyone has more details, please share.
Congratulations to the team!
Congratulations to the team!
That is great news.
ReplyDeleteSyriac MSS seem to be clearer than Greek MSS, to me. That said, I was trained to read Syriac MSS and not Greek ones, though I read Greek, too. There are divisions between the Syriac words, which makes it easier.
There is another project that accomplishes that task now (see http://kraken.re/) by using a neural network that can be trained to recognize and read handwritten scripts. It is a pretty cool design, even if I would not use it to produce a critical edition.
ReplyDelete