Book People Archive

FineReader OCR



Has anyone else tried FineReader 5.0 OCR
(http://www.abbyy.com/products/fine/index.htm)? It's available for an
on-line download for $99.00. I was at the Cornell University Library
last Friday, and an expert on library digital collections recommended
that I take a look at it. He said that the best OCR work was Russian
(although the Cornell Library has mostly used ScanSoft's TextBridge for
their OCR work).

I downloaded it and ran a quick comparison on some somewhat skewed,
medium quality scans. Here's a sample of the recognized output (of a
footnote) from FineReader - it's completely correct:

* The venerable Timothy Hatherly, although in Scituate in 1634, being
unmarried, had no house here until 1637, when he erected one on "farm
neck" within the Conihasset grant.

Now here's the output of the same line from Scansoft's TextBridge
Millenium:

* The ve,,erable 'Jlrnot)iy Halherly, although in Scitoate 111 :634
beh:g
unmarried, had no house here until :637, when he erected one on "tarn:
neck" within the Conihasset grant.

Seems like a pretty impressive difference to me. If anyone else has used
FineReader - or compared it to other OCR systems - I'd love to hear
about your experiences.

-- Dean Krafft

Dean B. Krafft
Director of Computing Facilities
Department of Computer Science
Cornell University
dean@[redacted]