**YouTube** is an American [online video-sharing](https://en.wikipedia.org/wiki/Video_hosting_service "Video hosting service") [platform](https://en.wikipedia.org/wiki/Web_platform "Web platform") headquartered in [San Bruno, California](https://en.wikipedia.org/wiki/San_Bruno,_California "San Bruno, California")

cf3f59da-c081-4fab-99d2-f02e60e183b0.PNG

The Extract semi-structured data activity allows RPA developers to easily take advantage of UiPath's machine learning models for semi-structured documents processing. The models currently can analyze invoices and receipts, providing various information (total price, vendor address, list of items, etc.). By simply specifying the image to process and the address of the model to use, the activity returns the model's response already formatted for parsing or verification using UiPath's Validation Station.

A typical use case would be to retrieve the total and vendor name from a list of scanned receipts and input them in an expense managing software (e.g. Concur). An advanced version of this process could even check for confidence scores and trigger a different workflow which will present validation station to a human user.

There are currently two machine learning models that can be targeted by the activity:

*   Invoices: https://invoices.uipath.com
*   Receipts: https://receipts.uipath.com

**You can find example scenarios in** [**this repository**](https://github.com/uipath-antoine/ExtractSemiStructuredDocumentExamples)[**: https://github.com/uipath-antoine/ExtractSemiStructuredDocumentExamples**](https://github.com/uipath-antoine/ExtractSemiStructuredDocumentExamples)

A custom activity to query UiPath's machine learning models for semi-structured document data extraction

Extract Semi Structured Document Activity