Hamburg score

We recently released ABBYY FlexiCapture for Invoices, a solution for streaming European accounts into the SED (Electronic Document Management System). Strictly speaking, the “big” ABBYY FlexiCapture is able to handle invoices, on the basis of which a new decision is made. But it is necessary to independently set all the settings for working with accounts in it, and many of our clients, as it turned out, prefer already customized solutions. Well, you understand - get the program out of the box, install Mary on the computer (the product is sold only in Europe so far), connect to the SED - and let it work.



The solution was developed specifically for processing only one type of documents - invoices. This made it possible to improve and “sharpen” the technologies specifically for these documents (although it allows processing other documents).
')
The most important thing - in ABBYY FlexiCapture for Invoices there is a universal flexible description with which you can correctly extract information from accounts that come from different countries. The program “understands” which country the account is from and “knows” which standards for billing exist in this country. This makes it possible to correctly determine where in the invoice, for example, the name of the supplier company, and where the total amount of the invoice is indicated.



The solution is based on advanced “flexible description” technologies created on a large sample of invoices from various suppliers. In other words, we took a lot of accounts from different countries and taught the program not only to distinguish them and correctly determine where and what they indicated, but also to decide whether it is possible to accept the received document at all.

We have written here in detail about how flexible descriptions work, so let's say briefly: A flexible description is a formalized description of the structure of a weakly structured document that is created using the FlexiLayout Studio tool. Unlike rigid templates, where the structure of all documents of this type is the same, for weakly structured documents the arrangement of fields may differ quite strongly. Such a flexible description of the field is looking for, relying on the heading keywords, on other fields that have already been found, and so on.

Each country has its own specific billing features: tax and legal rules, data formats, field locations, and more. For example, in Switzerland, when calculating taxes on invoices, their value is rounded to 0.05, often downwards. This is due to the fact that the smallest coin in Switzerland is 5 centimes (Santim - 0.01 from the Swiss franc). And, for example, in Spain, the name of the same company can be written in six different ways, and all these are standard abbreviations.

Such differences, on the one hand, somewhat complicate the task of creating a universal flexible description, and on the other hand, they help. And they help when searching for the remaining fields of the account. But before you search and recognize fields, you first need to determine the recognition language. And for this, you first need to determine the country from which the bill came. This helps the database that the user connected. The database contains a list of suppliers and the necessary information about them: name, address, bank details and, most importantly, the country. Having determined the country, the program looks at what settings are inherent in the accounts of this country (taxes, designations, reductions, etc.), then the remaining fields are searched for. This is how the flexible description is imposed and the necessary data is extracted.



Although the solution is ready for use, the user can improve and refine it to fit his goals, documents and business processes. FlexiCapture for Invoices uses a training technology that allows both the program administrator and the document verification operator (Mary operator - remember?) To adjust the flexible description.

In order to refine the universal flexible description taking into account the characteristics of documents of a particular supplier, the operator should not possess any special knowledge and skills. Refinement of the flexible description is in the processing of documents: if the operator noticed that any field was not found on the document, or was determined incorrectly, he changes the frame around the field with the mouse, presses the Train button. Pressing the Train button gives the program a signal that, for a given supplier, this field is in a certain place of the account, and the program remembers this, taking into account such factors as the field contents, its immediate surroundings, the geometric relationships of this field and other objects in the image, and other . Further, the flexible description changes to reflect new properties, and a new version of the flexible description is formed specifically for this supplier. The following documents from this supplier will be processed taking into account the new properties of the flexible description.





The result of training (the way it looks inside the program) is a flexible description consisting of several options (branches), each of which describes its subtype of a document from a specific supplier. In the process of learning, the program “binds” new properties obtained in the process of learning with a specific subtype of the document. The result of the imposition of a general flexible description is compared with the result of the trained variant, and the best result is selected for each individual field.

The program administrator has several more options to refine the template than Mary has. The whole list of options for a flexible description is available to him, and he can also teach each of these options. The name of each option is the name of the supplier company. "Inside" each - a sample of documents on the basis of which the training took place.



If the administrator decides that the software from a particular supplier is already being processed by the program quite well, he can disable the ability of operators to learn documents from this supplier so that Mary does not accidentally spoil the learning outcomes.
In addition to automatic training, the program has many opportunities for customization to the needs of a particular company. The program administrator has access to a list of all countries for which a flexible description has been configured. Each country has its own list of settings: a set of languages, legal and tax rules, a set and arrangement of fields. All these settings and properties can be edited (after all, tax laws sometimes change, even in Europe :)), as well as add new countries that are not in the standard list, specify their own settings for them.



A recognized and verified document can be exported with one of three flags (statuses): Valid - documents that successfully passed all checks, Exception - documents that require verification (checks), and Rejected - documents that have critical errors with which the document has been checked. cannot be accepted Documents with different statuses can be exported with different settings. The program automatically issues statuses as a result of checks, and the operator can change the status of the document, and he must write a comment to each document - the reason for changing the status.



If such an arrangement does not suit the administrator, for example, from the point of view of the company's business processes, he may prohibit the operator to change the value of the “Status” field.

Here is such a product. We hope, ABBYY FlexiCapture for Invoices will appeal to our European users.

Anastasia Evstratova
Corporate Projects Department