followers 0 popularity
2
following 1

following  view all

Groups
mahui is not in any groups
mahui

Supersize Your Strength

Jun 24th 2012 at 8:35 PM

This type of system combines OCR (Optical Character Recognition) technology and Contextual logic searches the keywords within the document and assigns these data in the document title. For example, if the word "date" find the data, it is usually closest to a Date field. With this knowledge, you can assume that the likely format and possible values ​​for the field that you just recorded and can automatically discount some possible OCR results (such as a monthly value will never be over 12). By combining technologies such as artificial intelligence Supersize Your Strength , several OCR modules and contextual and fuzzy logic, these systems give better results than the use of solid OCR templates - some manufacturers claim character recognition rates> 90%. You have to buy expensive and often complex to set up and refine. Many use a knowledge base that will be taught to recognize the need to type the invoices you receive. No matter how much effort you spend implementing your document management software, no system is 100% correct and faults are always crawling with all kinds of factors: dirty, wrinkled or torn paper, changes invoice layouts or even scanner errors.

OCR is the law of diminishing
As a document management software system will give 100% recognition rates, double check all your results or accept that the odd one will slip through the net. Unfortunately, there is the odd slip through the net cost of money at the track. It makes sense, as high as possible to seek a detection rate but this type of approach has a major drawback, it follows a law of diminishing returns is OCR. The more time and money you spend trying, OCR, performance improvement, the smaller the increase in recognition.

If your document management software, an 80% detection rate is achieved, you have to spend £ 5000 a 5% improvement achieved. Once you get to 90%, the same 5% will cost £ 50,000.

0 comments
Please to comment

sign in

Username
Password
Remember Me


New to IM faceplate? join free!

Lost Password? click here