Microsoft Azure Document Intelligence Neural vs. Template Models

In this session we will take a 12 document sample set, all with varying document types and formats, some handwritten correspondence, standard text-based forms, mortgage, insurance claims, medical claims and the like. We train 5 samples of the same document and then create a template and neural model in Azure Document Intelligence studio. We then test the same document set with both models. There is a clear winner, where neural model works great, makes little to no mistakes on the extracted data. Azure is able to recognize the trained document and only extract relevant fields. In the template portion of the demo, we see Azure pickup erroneous data fields that are not present on the document. The video will show some examples of fields that should not have been extracted.