GScan's simplest way of creating templates for automatic data extraction has become even simpler. In this post, you will find out, how the Learning mode feature evolved over the course of time from the initial implementation to the latest release and what does it currently offer.
23. 7. 2021 Tomáš Jurik
Learning mode is part of the GScan and GScan Service products for a long time now. It was designed to facilitate the creation of simple templates for data extraction from documents during the verification process. The idea was to allow the template creation to take place during the document processing itself and not needing to have them prepared beforehand.
The first iterations of the Learning mode feature were more like a teaching mode. Teaching mode in the sense, that the user had to "teach" the GScan, how to recognize the given document, and where to look for the required values in the document. The system was not really learning anything, as the user was providing all necessary information, therefore the template was user-created. It was a start.
The next iteration has become some kind of hybrid between teaching and learning. The system was automatically learning how the documents look and the user just had to define, where the data will be extracted from. This iteration was in production for a brief time only.
The current Learning mode iteration has finally earned its adjective - learning. The user is not anymore involved in the creation of recognition templates in this mode. GScan now learns automatically how to identify the document and also where to extract the data from. Once the document is imported/scanned, GScan creates a so-called document fingerprint. The document fingerprint serves, as the name suggests, for document identification. It is then used to compare it with document fingerprints attached to existing learning templates.
If a match is found, the corresponding template is applied, data extracted, and put into indexing fields.
If no match is found, no template is applied. In this case, the indexing fields are empty when the document is opened for verification. The data should be now extracted using the Click-to-Index feature (simply clicking on the required values in the document preview). GScan learns during this process, where to extract the data from. Based on this information, GScan creates the template, attaches the document fingerprint to it and it is saved. All this without further user involvement. The next time such document is imported/scanned, its fingerprint is again compared with fingerprints in all existing templates, and the cycle continues.
Request a free trial or a quote. Ask us anything via chat or leave us a message in email.
