Hi,
I recently built a machine learning model to perform Entity Extraction from documents by using Google AutoML connected system. It does not return expected result as compared to Postman, when Predict operation was invoked from integration service. Can someone give me pointer on what should be the input format to be used to get proper result.
Note - I used Google OCR to extract document detail into plaint text. This was then fed into Predict Operation in Appian as well as Google AutoML RESTful API to predict document entities.
Following are few approaches i tried from Appian integration service to produce intended output.
Approach - 1
Data Set Type - Text,
Text Input -
"Driver Licence\nNew South Wales, Australia\nDavid Macaulay FISHER\nCard Number\n2000 508 479\n29 ABC ST\nASHFORD NSW 2361\nLicence No.\n12345678\nLicence Class\nC\nGayelord Phil McBride\nDate of Birth\n01 JAN 1950\nExpiry Date\n20 FEB 2023\n"
Output
Approach - 2
"{ "payload": { "textSnippet": { "content": "Driver Licence\nNew South Wales, Australia\nDavid Macaulay FISHER\nCard Number\n2000 508 479\n29 ABC ST\nASHFORD NSW 2361\nLicence No.\n12345678\nLicence Class\nC\nGayelord Phil McBride\nDate of Birth\n01 JAN 1950\nExpiry Date\n20 FEB 2023\n", "mime_type": "text/plain" } } }"
Approach - 3
Data Set Type - Image
Postman Output
Discussion posts and replies are publicly visible