Converting an image into a searchable PDF file becomes the standard. However, this process is often fragmented (scanner-based OCR, the need for using additional applications, etc.). Within our model, the OCR service operates at the server-side, is efficient because of its container architecture (docker), and ensures that each document will be in a searchable form.

Documents in text format and information entered on the form (metadata) can be searched contextually (as in Google) using specialized tools: Elastic search, Full-text search, etc.

General data capture models are ready to use without additional training. Most customers use the model for sales documents (invoices), but we are working on others (business cards, CVs, etc.).

You can either train general models or create your own from scratch. While model training (e.g., for invoices), we adjust the existing field settings or add recognition of new ones. You can save settings for a specific contractor. You can also design models for new document classes.

The primary use of text classification is to identify the document type. Based on the document content, the system evaluates whether it is an invoice, a letter, a contract, etc. The trained type classifier is individual for each environment.

Another use of the text classifier is to estimate values for a particular field. The system recommends, for example, a document holder, accounts, or a cost center based on its analysis of historical documents.