Melissa Terras is the great expert on this, of course, along with our own Arianna Ciula and Paul Stokes. My understanding from a talk I heard her give last year in London is that it is a holy grail of OCR companies, but nowhere near practical.

I'm not an expert on the current state of this field, but I believe the approach currently is at best to assist humans in the interpretation of characters by looking for patterns that can be highlighted, rather than attempt to replace humans in interpret strings of manuscript text. So in this regard something like how we used to use photocopiers to increase contrast or different types of light to emphasis different types of strokes or ink. But now going farther than this in looking for patterns that can then be approved or interpreted by a human.

So my sense is that if the question is "can I use a computer to avoid keying the content of this manuscript into a computer?" the answer is no. But if the question is "can I use a computer to help me recognise, interpret, or classify aspects or characters in this script?" the answer is yes depending on the specific issues you are looking at analysing.

Good convenient starting places on the current state of the field are the articles in DM by Paul Stokes (newly published in DM 3: http://www.digitalmedievalist.org/journal/3/stokes/) and Arianna Ciula (DM 1: http://www.digitalmedievalist.org/journal/1.1/ciula/)%5B1]. We have a review coming of Terras's book _Image to Interpretation: An intelligent system to aid historians in reading the Vindolanda texts_ (Oxford: 2006).

-dan

[1] If you haven't been to the DM site in a while, you'll see it has been improved and reorganised. Old URLs for journal articles should still work however.

On Fri, 2008-03-21 at 13:58 -0400, Dot Porter wrote:

Dear List,

I've received a query from a medievalist who is interested in applying OCR to manuscripts. I'm not really aware of recent work in this area and I'm wondering what, if anything, is being done at this time or in the recent past. Last time I looked into it good OCR from handwritten texts was a long way off - for nicely written, straight English text, to say nothing of heavily abbreviated medieval Latin or Old English writing. But I'd be delighted to be proven wrong.

Thanks! Dot

--

---

Dot Porter, University of Kentucky ##### Program Coordinator Collaboratory for Research in Computing for Humanities http://www.rch.uky.edu Center for Visualization and Virtual Environments http://www.vis.uky.edu dporter@uky.edu 859-257-1257 x.82115

---