Digitization of documents by data extraction using AI

The Task

Slovenská sporiteľňa, a.s. (hereinafter referred to as SLSP) in its Retail Process Center (hereinafter referred to as RPC) processes incoming paper correspondence for the entire branch network.

In order to digitize documents and at the same time reduce the duration of the process of transcribing data from paper documents into the information system, SLSP approached us with a request to create a solution Concept (PoC) of automation of text extraction and processing from various types of scanned documents. In the document "Consumer Payment Account Information", our task was to identify and extract the required client data, including information on direct debits and standing orders on the account, and provide these in a defined format for transmission to the target information system.

From SLPS's point of view, automation had to meet KPIs:

• Shortening the time for processing the agenda to less than 17 minutes
• Accuracy of data recovery at the level of 90%.

The solution

Our application for intelligent data extraction DocMiner was customized for client purposes so that after processing the document using OCR technology (ABBYY FineReader) and converting the text into a data form, based on defined algorithms, criteria, and rules, it identifies the necessary data, validates and aggregates it into a data sentence required format. At the same time, DocMiner enabled mass processing, i.e., processing of several documents at once, automatically, without manual execution.

The result

The delivered solution enabled the mass processing of dozens of documents with the resulting success of data extraction at the level of 97.26% per individual extracted item. The processing time was shortened according to the client's requirements.