IssueTracker

Problem you have encountered:

Customers dealing with multipage documents often experience time-consuming and expensive error correction due to a lack of specific error location data (like page number). Currently, errors of the kind [1] only indicate the entity type, making it difficult to pinpoint the problem within large documents.

Adding page number (and potentially line number) to the error metadata would significantly streamline the correction process. This would allow users to quickly locate and resolve labeling errors within multipage documents.

Customer requested if this improvement can be prioritized as they deal with these kind of situation in hundreds of documents when their training intents fail.

What you expected to happen:

Get the page number in the error object metadata to be able to speed up the label correction issue.

Seen error:

[1]

{
    "code": 3, 
    "message": "Invalid document.",
    "details": [
       {
         "@type": "type.googleapis.com/google.rpc.ErrorInfo",
         "reason": "INVALID_DOCUMENT",
         "domain": "documentai.googleapis.com",
         "metadata": {
             "num_fields_needed": "1",
             "num_fields": "2",
             "document": "gs://my_bucket/my_file.pdf",
             "annotation_name": "my_annotation_name",
             "field_name": "entities.page_anchor.page_refs"
           }
        }