HN Explorer

8 comments

[ 7.7 ms ] story [ 38.6 ms ] thread
appleflaxen
This is just an advert with very little substance.
eastendguy
Yeah, nothing to learn from this page. But I can not undo my upvote. I was hopping to see either a comparison of established services (Abbyy, Google Cloud Vision, OCR.space,...) or, even better, an insightful report from a Tesseract developer.
AndrewKemendo
That would be very helpful.

Anecdotally, I found ABBYY to blow everything away, followed by the cloud provider services led by MSFT and then OSS stuff led by Tesseract.

piterdevries
"Dark data" almost as catchy as "Dark Matter"
vavooom
Does anyone have any resources for 'Build it Yourself' form of OCR software. I am aware of Python packages like pytesseract that are somewhat customizable, and am curious if other, pre-built platforms are available.
kumartanmay
It depends from one form to other and sensitivity of data. Our company has developed a product to transform data from pdfs (electronic & non-machine readable files) for financial services. To meet the needs of customers, we have developed parsers for bank statements, loan/lease agreements, valuation reports and portfolio analysis. There are 20% of the documents that are scanned and require OCRs to be trained for different fonts. We wrote about challenges with currents OCRs: https://medium.com/zodhana/modern-day-challenges-with-docume...
eastendguy
Tesseract is the only open-source OCR package that I am aware of. If anyone knows anything else, I would be very curious as well.