I am running some example custom processors to read values of governmental documents. Some of the values we are after have a single letter as value, for example on a passport "Gender" has a value F or M. With a very high precision during labeling, this letter cannot be found by the tool provided. The labeling teams added the value manual but then the training of the processor fails with:
"code": 3,
"message": "Invalid document.",
"details": [
{
"@type": "type.googleapis.com/google.rpc.ErrorInfo",
"reason": "INVALID_DOCUMENT",
"domain": "documentai.googleapis.com",
"metadata": {
"document": "XXX/61275_D88_FIN_46.JPG",
"num_fields_needed": "1",
"field_name": "entities.text_anchor.text_segments",
"annotation_name": "Gender",
"num_fields": "0"
}
How to work with such cases as we have in many documents which single letter values we like to parse?