 |
 |
Hello Wendy, long time no speak. I've worked with OCR software many times over the last few years - it's fair to say that it can give mixed results. OCR cannot effectively read hand-writing - yes there have been projects to read handwritten records, but these have been typically based around standard response, handwriting in capitals or forms, where ticks in boxes can be recognised. OCR will sometimes read records perfectly, sometimes, simply depending on the typeface it will produce complete rubbish. The other problem you can get, especially with older type-faces is that the software can read l as 1, m as ni or S as 5, etc. The software will also often sprinkle the document with capitals, reading an s as an S, etc. Whilst a spell-checker will sort some of these out, sometimes the change can go unnoticed as a it creates a real word - also if you are using scientific terms or latin names, a spell-checker will be useless anyway. Finally, OCR software can give variable results with columns. There are specialist providers out there, but this can be expensive. Don't think I am being negative - more ....realistic. OCR works best where the input documents are simple, predictable and consistent. Where there is variation or anything unusual, it starts to run into problems. Even the best OCR software will only ever be 95%-99% accurate - this sounds good, but it could mean one in every 20 characters is wrong. The end result will be a lot more manual input to check and correct the digitised data - this to a degree defeats the whole object of OCR'ing in the first pace. Sometimes it is cheaper and easier to hire in a professional typist to bash the details in. The rate of throughput, accuracy and cost can often give OCR a serious run for its money. Regards Chris Meaney (AIMC) Managing Director ======================================================================== Harvard Consultancy Services Ltd, Bexin House, 2/3 St. Andrews Place Southover Road, Lewes, East Sussex, BN7 1UP Tel: 01273 897517, Fax: 01273 471929, E-Mail: [log in to unmask] Registered in England & Wales no. 3766540 Registered Office: 50 Harvard Close, Malling, Lewes, East Sussex, BN7 2EJ. -----Original Message----- From: Museums Computer Group [mailto:[log in to unmask]]On Behalf Of Ian Morrison Sent: 20 February 2002 09:00 To: [log in to unmask] Subject: Re: Advice on OCR software please On Wed, 20 Feb 2002, Wendy Sudbury wrote: > Could colleagues share their experience of successful use of OCR > software? We have A5 index cards, typed in assorted fonts, 4 or 5 > columns across, with a few handwritten markings (numbering and ticks in > pencil or biro). I'm looking to migrate the information to a table in > Word whenever we have to update a card. We update modest quantities, > maybe 40 a month. It seems to make sense to scan the old card first. I have done some experiments with various kinds of museum records over the years, latterly using an evaluation version of PageGenie 98 with some success. I don't know if it is still available for free download but, if so, it might be worth giving it a try. The success will probably depend on the consistency of the original typing - electronic typewriters generally give the best results. So far I haven't found any OCR software that will make much sense of the average curator's handwriting, sadly, nor do I know of any museum that has used OCR successfully to digitise records in any numbers. Hope this helps. ------------------------------------------------------------- Ian O. Morrison, Scottish Museums Documentation Officer http://ianmorrison.topcities.com/index.htm Hostes alienigeni me abduxerunt