anderson@uwmacc.UUCP (06/15/87)
I have a large, handwritten document that I would like to read with a scanner and convert the original text to Ascii text. I suppose the hardest part of the problem is teaching the program recognize the several signs that can represent any given letter. The expert system itself doesn't seem to me like the daunting part. The hardest part would be recognizing the signs, since I know nothing about that problem theoretically. The document is big, you see -- over 37,000 pages -- and even counting in the time it would take me to develop the software, there would be a likely saving over the time it would take to enter the text by keying it in. The original took 14 years to write. I'd appreciate any pointers anyone had to offer. -- ==ARPA:===============anderson@vms.macc.wisc.edu===Jess Anderson====== | UUCP: {harvard,seismo,rutgers, (avoid ihnp4!) 1210 W. Dayton | | akgua,allegra,usbvax}!uwvax!uwwircs!anderson Madison, WI 53706 | ==BITNET:======================anderson@wiscmacc===608/263-6988=======
sarah@laticorp.UUCP (Sarah Groves Hobart) (06/16/87)
In article <1633@uwmacc.UUCP> anderson@uwmacc.UUCP (Jess Anderson) writes: >I have a large, handwritten document that I would like to read >with a scanner and convert the original text to Ascii text. You don't say if your handwritten document is in cursive or hand-printed form. If it's hand-printed, a good OCR could do most of your work for you. Some OCR's will use context to improve character recognition, using known patterns of English characters to determine what the suspicious character could logically be. (Is your document in English?) Even if the document is in cursive, take some sample sheets of the document to your friendly OCR salesperson and see if any of the commercial scanners can handle it as is. A lot of scanners will make a valiant try, and the error rate could be acceptable if you're going to edit the document anyway. Look for ones that will conveniently mark uncertain characters and words for you. (Now I may be talking through my hat here . . .I've never fed cursive handwriting through a scanner, but a lot of them say they will try to separate characters correctly.) In a month or so I'll be working with a Palantir OCR, so I'll be interested to see what your request turns up. If you get much interesting mail on this topic, please post a summary. Thanks! Sarah Groves Hobart