We need to process invoices with variable structures. The question is: these invoices can be in many languages and come from different countries. Specific fields are required for each country.
Do we need to create a template / document type for each one of the countries? Do we use just one ML model and train it in the different languages + different fields or we need specific ML models for each country?
It is possible to use only one ML model.
Try to create Semi-structure AI and use invoice as out of the box document type.
There is info how many document is recommended for specific number of languages, fields and layouts.
I am more or less familiar with the DU framework and the official documentation.
I was asking about what you meant with “Semi-structure AI” (you mean a ML skill?) and how the package would differentiate between countries. I am probably mistaken but Im not easily finding info about it
This semi-structured AI is an extraction method used when you create new document type.
(Go to document understanding tab in orchestrator, create new project and select new document type with option ‘using semi-structured AI’). You can choose one of existing pre-trained model (e.g. for invoices) or select custom document type. It is possible to define additional custom fields or columns. This tool is good for strengthening UiPath pre-built models on your specific layouts