Form Recognizer / DocumentIntelligence Studio JSON schemas missing

Joe Nudell 30 Reputation points
2024-04-29T15:08:35.33+00:00

The JSON schema URL referenced in the *.labels.json files generated by the Document Intelligence studio is no longer valid: https://schema.cognitiveservices.azure.com/formrecognizer/2021-03-01/labels.json

We use these labels.json files in our model evaluation code in order to find the bounding regions of the labels on our hand-labeled documents. Our code tries to parse these files according to the $schema provided. This code is now broken since the schema seems to have disappeared / the URL is broken.

The $schema URLs in all of our fields.json and *.labels.json documents in our blob storage are now broken. This seems like a bug? Can the schemas be restored at the specified URLs?

Azure Document Intelligence in Foundry Tools

Answer accepted by question author

YutongTie-9091 54,026 Reputation points Moderator
2024-04-29T20:03:31.69+00:00

@Joe Nudell

Thanks for reaching out to us, based on the note from the product team, this content is under maintenance.

I am sharing the content below for your reference, the new external website will be published soon.

labels_schema.json

{
 
"title"
"type"
"properties"
"document"
"description"
"type"
 },
"labels"
"description"
"type"
"items"
 }
 },
"required"
"definitions"
"Label"
"description"
"type"
"properties"
"label"
"description"
"type"
"format"
 },
"key"
"description"
"type"
"null"
"string"
 ]
 },
"value"
"description"
"type"
"items"
 },
"labelType"
"description"
"type"
"enum"
 }
 },
"required"
 },
"Value"
"description"
"type"
"properties"
"boundingBoxes"
"description"
"type"
"items"
 },
"pageNumber"
"description"
"type"
"minimum"
 },
"text"
"description"
"type"
 }
 },
"required"
 },
"BoundingBox"
"description"
"type"
"items"
"type"
"minimum"
"maximum"
 },
"maxItems"
"minItems"
 }
 }
}

fields_schema.json

{
 
"title"
"type"
"properties"
"fields"
"description"
"type"
"items"
 },
"definitions"
"description"
"type"
"patternProperties"
// Only allow names that do not mach one of the predefined field types.
// In #/definitions/, the fieldKey is specified as part of the dictionary.
"^(?!(string$|number$|date$|time$|object$|array$|integer$|selectionMark$|checkbox$))"
 },
"additionalProperties"
 }
 },
"required"
"definitions"
"FieldWithoutKey"
"oneOf"
 { 
 { 
 { 
 { 
 ]
 },
"Field"
"allOf"
 {
"properties"
"fieldKey"
"description"
"type"
 }
 },
"required"
 },
 { 
 ]
 },
"PrimitiveField"
"description"
"type"
"properties"
"fieldType"
"description"
"type"
"enum"
 },
"fieldFormat"
"$ref"
 }
 },
"required"
 },
"FieldFormat"
"description"
"type"
"enum"
"default"
 },
"ObjectField"
"description"
"type"
"properties"
"fieldType"
"description"
"type"
"enum"
 },
"fields"
"description"
"type"
"items"
 },
"visualizationHint"
"description"
"type"
"enum"
 }
 },
"required"
 },
"ArrayField"
"description"
"type"
"properties"
"fieldType"
"description"
"type"
"enum"
 },
"itemType"
"description"
"type"
 },
"visualizationHint"
"description"
"type"
"enum"
 }
 },
"required"
 },
"CustomField"
"description"
"type"
"properties"
"fieldType"
"description"
"type"
"pattern"
 }
 },
"required"
 }
 }
}


Sorry the format here looks weird, please let me know if you need an email dropped, I can share the files to you.

I hope this helps.

Regards,

Yutong

Was this answer helpful?


0 additional answers

Sort by: Most helpful

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.