LAD/RAD: Automatic Document Reading and Recognition
LAD for automatic document reading, RAD for automatic document recognition: in most cases, these 2 technologies are combined and work together. The growth of LAD/RAD solutions is closely linked to that of dematerialization. The company receives paper documents – invoices, faxes, letters, etc. – which it needs to digitize in order to quickly process key information and enrich its information system. The incoming paper is scanned, and the RAD/LAD software reads and retrieves the information. These can then be integrated into the EDM system or used to automate document processing.
By definition, LAD/RAD brings together the technologies needed to categorize documents and capture textual information from structured or semi-structured digitized documents.
- Software uses ARD to automatically recognize and categorize documents The technology compares the scanned document with templates in its database, classifying it as a quotation, an ID document, an order form, etc. The classification criteria are varied, and may include a supplier’s logo or the word “ex VAT”, for example.
- LAD technology then identifies the key information to be extracted: the data from the scanned paper document is then captured, so that it can be easily identified and exploited in digital format.
From readable document to comprehensible document: the different generations of LAD / RAD
OCR, LAD / RAD, Smart OCR, Smart Capture, Document AI… When it comes to document capture, there are so many terms that it’s sometimes difficult to make sense of all the different solutions.
Let’s cut to the chase: LAD / RAD, OCR, Smart OCR and Document AI all mean the same thing. These different processes all refer to the fact of being able to read a document in order to automate its processing. It’s just that the technology used to achieve this differs from one generation to the next, which has a major impact on the solution’s performance.
By analogy, all vacuum cleaners have the same function: to suck up dust. But there’s a big gap between the first vacuum cleaner, which used manual bellows, and the latest Dyson. Well, it’s exactly the same between the different generations of LAD / RAD. Depending on whether they use a rule-based system, standard Machine Learning or Deep Learning, the results change dramatically!
First-generation ADSL: the beginnings of document reading
Historical ADL relies solely on OCR (optical character recognition), which transforms an image into text.
Second-generation ADL / ARD: a tangle of rules
A sub-category of Machine Learning, Deep Learning is based on deep neural algorithms that deliver prodigious performance. When LAD / RAD uses Deep Learning, there’s no need to set any parameters – the machine will make up its own mind.We don’t have to explain or conceptualize the difference between one type of document and another. No need for rules, the machine just needs examples!
To better understand this, let’s imagine we ask the machine to recognize a cat from a dog, from one image to another. While a 6-month-old child might be able to perform this task, it’s virtually impossible for a rule-based system. Cats and dogs have 4 legs and a tail, they can be small, with long or short hair, with larger or smaller ears… Their identification is based on a combination of differences, immediately perceptible to a child, as to a Deep Learning model. In this case, Deep’s model simply needs to be trained on a series of images of cats and dogs, so that it can then recognize them on new images it has never seen before.
What’s more, with the new-generation LAD / RAD, it is now possible to perform LAD and RAD at the same time, resulting in automatic processing of the complete document. Indeed, a Deep Learning model pre-trained on several documents will both recognize it and extract what it needs to extract. This is a major advantage in terms of workflow simplification and solution performance.
- When LAD / RAD uses Deep Learning models, we speak of Smart Capture, Document AI or Intelligent Document Processing.
reciTAL is the first Deep Tech-certified LAD/RAD company to publish automatic document and e-mail processing software, enhanced by the latest AI technologies. The ultimate Document Processing Intelligence solution, reciTAL speeds up document categorization and information capture, making them more reliable and helping to boost the performance of your solutions. Want to know more? Request a demo!
