What are the new "Build Mode" options in Form Recognizer?

Question

Hi,

I recently noticed that there were some new options available when training a custom model in Form Recognizer. I was wondering if there was some information on what each of these options are, as currently it is unclear as to which one I should be using.

The two options appear just after you have tagged your documents and click the blue "Train" button. It asks for a model ID, model description, and now also a Build mode. This is a fairly new feature, as I was just in here training documents last week and did not see this. The two options within the build mode dropdown are "Neural" and "Template". However I cannot find any nformation available to detail what each one means.

I clicked on the "Help" that is advertised on the page for "Neural", however this takes me to an Error 404. I have tried looking this up on the web and on this forum, however there doesnt seem to be anything which can help me identify what this is.

Any help would be greatly appreciated

Thanks,
Yasmin

Answer 1

Hi, here's the information I found:

• Template models work well when the target documents share a common visual layout. Training only takes a few minutes, and more than 100 languages are supported.
  
• Neural models can flexibly handle both structured and unstructured documents. Training takes up to half an hour, and currently only English language documents are supported. The current version can extract inline field data and checkboxes. Neural models are available only in select regions.
  

I have informed the product team regarding lack of official documentation. Hope this helps!