Error Reporting#
We recommend using the feedback-api for error reporting.
Feedback-API#
The POST /processing/feedback/{processing_id} endpoint is available to allow the reporting of erroneous extractions or classifications.
The processed document will be added as training data to the corresponding workflow for annotation.
Parameters#
| Field | Type | Description |
|---|---|---|
{processing_id} |
UUID | The UUID of the processing you want to report an error for. |
Request body#
| Field | Type | Description |
|---|---|---|
description(optional) |
string | Permits describing the error in a plain and explicit manner. This information will be shown during annotation, so that you can easily fix the error. |
tag(optional) |
string |
This is only supported for fine-tuned or custom extractions, not for custom document classifications. You can specify the 'tag' of the training data. The tag refers to the name of the issuer/template you created when adding data to our interfaces. The erroneous document will then be placed in this bucket so that you can keep your data sorted and categorised. If you do not use the 'tag' parameter, the erroneous document will be placed in the unknown bucket. |
expected_class(optional) |
string |
This field is only relevant for custom document classification, not for fine-tuned or custom extractions. The document class that was expected for this file, if known. When provided and the provided class is supported by the related classification workflow, then the new file will automatically be annotated with this class and forwarded to review. If not provided or set to null, the new file will be added in unannotated state. If provided and the given class is not supported by the related classification workflow, it will be rejected and cause an HTTP 400 error. |
Error-Reports API (old)#
Caution
Please note, that this API Endpoint needs some additional configuration. For that get in contact with us. We always recommend using the newer feedback API route instead.
The endpoint POST /processing/error-reports/{processing_result} is available to allow the reporting
of errors related to document processing and classifications. This endpoint provides not only the ability
to simply state an error but also provide any specific fields that may be incorrect as well as the correct
value of that field. This information can then be used to help improve our system.
Parameters#
| Field | Type | Description |
|---|---|---|
{processing_result} |
string |
Passed as part of the query path, defining the aspect of the document processing that contains errors.
Currently, only extractions is supported.
|
Request body#
The request body can define the description of the error, but also the specific fields that contain an error and what the correct value should be.
| Field | Type | Description |
|---|---|---|
processing_id |
UUID | Defines the process instance that needs correction. |
description(optional) |
string | Permits describing the error in a plain and explicit manner. |
incorrect_fields(optional) |
array |
Used to indicate the specific fields that are incorrect. Allows providing both the field, as well as the correct value, as a part of each entry.
Each entry follows the general form of:
{
"field": "customer.name",
"correct_value": "Customer's Name"
}
The dot notation, used for the field, defines the specific attribute of the entity.
|