.jpeg)
In France, companies are mostly confronted with non-standard documentary fraud (invoices, quotes, pay slips, CPAM certificates, etc.) which concerns today 65% of fraud cases. With an increasing volume of documents received by businesses, understanding reading and automatic processing technologies has become essential. Three acronyms come up often: LAD, RAD, and OCR. But what do they really mean? And what are their specificities? Here are the essentials to remember to master these technologies in your document strategy.
OCR (Optical Character Recognition): transform the image into text
THE OCR (Optical Character Recognition or optical character recognition) is one of the best known technologies.
Its role is simple: convert an image containing text into editable digital text. For example, it makes it possible to transform a scanned PDF or a document photo into text that can be used in a computer system.
What is the typical use of OCR?
Optical Character Recognition makes it possible to make a printed or scanned document usable in databases. Today, it is mainly used to extract accurate information from a panel of documents.
How does OCR work?
THE OCR (Optical Character Recognition) generally follows a fairly structured process. To ensure the most reliable reading possible, this technology is based on three main steps:
1- Pre-processing the document : before reading, the document is improved in order to optimize character recognition (adjustment, contrast improvement, format standardization, etc.).
2- Recognizing and converting text : the tool then analyzes the characters present on the document and interprets them using an algorithm capable of identifying their shape, then converting them into digital text.
3- Post-treatment and correction : a final phase makes it possible to refine the result by correcting any errors, in particular using dictionaries or business references, in order to obtain more coherent and usable data depending on the use case.
However, OCR is generally limited to printed characters: it recognizes typed letters and numbers, but it struggles with handwriting or very complex shapes without specific learning.
RAD (Automatic Document Recognition): recognize the type of document
The RAD (Automatic Document Recognition) does not focus directly on textual content, but on identifying the type of document.
It analyzes the visual structure of a scanned document and, using models or learning algorithms, determines whether it is, for example, an invoice, a contract, a pay slip or a form.
What is the typical use of RAD?
The primary objective of RAD is to automatically sort and classify documents according to their nature. It is particularly useful when you need to direct each document to the right treatment or to the right file without human intervention.
- Sorting documents according to their category (administrative, legal, medical, etc.) and their nature
- The application of specific rules for each category (validation, archiving, indexing, etc.)
How does RAD work?
RAD is able to distinguish an invoice from a pay slip, or a form from a contract.
Its operation is generally based on two key steps:
- Analysis of visual characteristics such as logos, text blocks, the position of elements on the document
- Semantic analysis thanks to the presence of keywords such as “certification”, “Amount” etc.
LAD (Automatic Document Reading): the complete automation of document reading
The LAD (Automatic reading of documents) is in turn a broader approach.
It often includes RAD and OCR but goes beyond that: it aims to automatically extract the relevant data contained in your documents, then to make them usable in your systems (GED, ERP, CRM, etc.).
What is the typical use of LAD?
The LAD has several goals.
First, it makes it possible to analyze the structure and content of documents.
Then, it classifies them according to their type (invoice, form, contract, expense report, pay slip, etc.).
Finally, it extracts the relevant information, such as name, date, date, address, company, amount, VAT, etc.).
- Automatic feeding of your business flows or databases
How does LAD work?
LAD is a combination of recognition and data extraction, used to automate document processing and verification as much as possible.
Concretely, it is based on several types of analyses:
- Identifying the type of document using RAD
- Reading characters via OCR
- Extracting and structuring critical information
And the ICR in all this?
Even though the question was about LAD, RAD, and OCR, it's useful to briefly mention ICR (Intelligent Character Recognition).
It is an advanced form of OCR that can recognize handwritten texts and learn new characters over time using artificial intelligence models.
ICR therefore complements OCR in cases where the documents contain information written by hand.
LAD, RAD and OCR: complementary tools
To summarize these different technologies, here is a summary table!
Each technology meets different needs, but they are often used together in dematerialization and Electronic Document Management (EDM) solutions in order to offer a complete capture of incoming documents.
Why are these technologies important for anti-document fraud?
What is the advantage of these technologies in the context of the fight against document fraud?
In the context of documentary anti-fraud, efficient automated reading is essential for:
- Detect inconsistencies in extracted data
- Identify falsified or modified documents
- Comparing extracted information to sources of truth
- Automate controls while ensuring high detection quality
- Compare extracted data to customer journey eligibility criteria
To find out more about document fraud, read on: The big figures of document fraud in businesses
How do we use OCR, LAD and RAD at Finovox?
At Finovox, documentary reading technologies such as OCR, LAD and RAD are not an end in themselves. They constitute the path to a wider process of detecting documentary fraud.
This first path is more commonly referred to as: the Document processing. Within Document Processing, Finovox distinguishes 4 stages:
- The pretreatment : it is the improvement of the overall quality of the file. The documents are prepared to be analyzed.
- Then, OCR makes it possible to convert the content into usable text
- While RAD automatically identifies and categorizes the type of document
- And that the LAD extracts the relevant fields (name, date, date, amount, IBAN, address, etc.) in order to structure them
Once this data is made usable, it feeds into the following steps of the Finovox solution: Validation document, Fraud analysis and Case Investigation.
In other words, OCR, LAD and RAD serve as the basis for more advanced controls, capable of verifying the authenticity of the document, detecting inconsistencies or manipulations, and facilitating the investigation in case of suspicion.
Do you want to strengthen your documentary verification processes? Discover how Finovox can secure your document management and fight effectively against fraud.
Sommaire
