The Document AI Scanner is an Oracle APEX 23.2 application that leverages the Oracle Cloud - Document Understanding AI Service in a modern APEX user interface. The aim of this application is to process several PDF or image files in one job and provide the results of the analysis as a ZIP file for download or export to an accounting database.
The application accepts a set of files and then passed them to the Document Understanding AI Service for processing. At the completion of the process, the application retrives the analysis results and stores them in sql tables.
The application uses a algorithm that improves the key-value results provided by the Document Understanding AI. This algorithm uses alias names, data type validation, and relative positions to form valid key-value pairs from text fragments with x,y-coordinates. It is extracting field values based on their position relative to the alias names in the text fragments.
Install the file Document AI Scanner App.sql as an APEX application in your Oracle Autonomous database. Also install the supporting objects in the installation process. There is decent documentation https://github.com/dstrack/Document-AI-Scanner/blob/main/docs/Document%20AI%20Scanner%20App%20-%20Application%20Guide.pdf Follow the detailed installation instructions in the "Document AI Scanner App - Application Guide.pdf" to enable the connections to the REST-API and the object store.
The app uses the standard model of the AI service at the API endpoint: https://document.aiservice.eu-frankfurt-1.oci.oraclecloud.com/20221109/processorJobs
In the app there is a configuration page with all parameters. When installing the supporting objects, the SQL skrips are installed from the repo. In the file Document_Scan_Ai_Field_Alias_DEU.sql you will find the German field aliases for the supported key-value pairs.
Contributions are welcome! Please read our contributing guidelines before getting started.
MIT