Issues related to no OCR results
Case 1: OCR returns empty ocr_json
OCR returns empty ocr_json due to failures on the OCR side. The empty ocr_json column causes issues on further steps when a bot tries to process the column.
Check the ocr_failed column. In case of an error, it should have the true value. Then, check the ocr_failed_response column. It should contain a message about the occurred error.
Case 2: OCR returns wrapped error page
If the issue occurs not with OCR but when a file is being picked up (for instance, you get access denied, file not found, or Minio error), the OCR service processes the error message as an HTML or XML page.
In this case, you can face issues with further ML model processing because documents won't contain valuable text. See the xml_document_link variable, and you will find that it's wrapped an OCRed page with an error message.
Case 3: OCR returns no response
If you send the request to OCR and get no response, proceed as described below:
Check that the service gets the request using OCR REST API:
/cloud/summary. You should see that the task gets to theQUEUEDqueue and then moves toINPROGRESSandCOMPLETED. Otherwise, you can assume it is stuck somewhere on the way to the OCR service. Also, check that the OCR endpoint is configured properly and OCR is accessible and running.If you see the task gets to the service, but you get no response, check OCR logs and the OCR Worker on Marathon and Mesos:
- For the ocr-rest service, the logs are located on the master server path:
/opt/workfusion/ocr/logs. - For the OCR worker, the logs are located on the agent server path:
/opt/workfusion/logs/workers/log/worker/bep-ocr-task/*.
- For the ocr-rest service, the logs are located on the master server path:
Read also: