OCR Results shows every OCR job on this server and lets you stop, restart, and re-queue jobs. Each row is one OCR job for one document. This report is only available on cloud hosted sites.



Search Criteria

Submitted - The date range in which the job was submitted to the OCR queue.


Completed - The date range in which the job finished.


Record ID / Document ID - Limit the report to the jobs for one record or one document. When you open this report from a record or document these are filled in for you and Not Complete Only is turned off so finished jobs are included.


OCR Mode - All modes, Standard OCR only, all Machine Learning modes, or one specific Machine Learning mode such as Invoice Processing or Receipt Processing.


Not Complete Only - Only jobs that have not finished, meaning they are waiting in the queue or currently processing. This is on by default.


Not Queued Only - Only jobs that never made it into the OCR queue: Status 0 with no status text at all. A job waiting its turn shows Queued and one handed to the processing pool shows Dispatched, so a job that still has no status a few minutes after it was submitted was lost and needs to be re-queued.


Has Errors Only - Only jobs that failed or were stopped. These are the jobs that can be re-queued in bulk. See Re-queueing Failed Jobs below.


OCR Tag - Enterprise only. Show the jobs that were started by a workflow or add-in action with this tag.


Results per Page - How many rows to show on each page.


Example Output

[Screenshot: OCRResultsReport-Results.png - retake with the check box column]

Rows for jobs that are still queued or processing refresh themselves every 10 seconds. Click a row to preview the record and document. The columns from left to right:


Check box - Shown only on jobs that failed or were stopped. Check the jobs you want to re-queue. See Re-queueing Failed Jobs below.


Submitted / Started - Submitted is the date and time the job was submitted to the queue. Started is the date and time the job actually started processing.


Status / Last Update - Status is the current state of the job. Queued means the job is waiting in the OCR queue for a processing slot, Dispatched means it has been handed to the processing pool and will start shortly, and Not Queued means the request never made it into the queue. A job that is processing shows its current step and a progress bar. A finished job shows Complete, or the error text if it failed or was stopped. Last Update is the last date and time the job changed. For a finished job this is when it finished.


Tag - Enterprise only. If the job was started by a workflow or add-in action with a tag, the tag shows here.


Rec / Doc / OCR - The IDs of the objects being processed. Rec is the ID of the record that held the document when the job started. Doc is the ID of the document being OCR'd. OCR is the ID of the job itself, which is mostly useful for back end troubleshooting.


Pgs - The number of pages in the document. This is 0 until the job starts processing.


ML - The OCR mode. Hover over the icon to see whether this is a standard OCR job or a specific Machine Learning mode.


WF - Whether the job was started by workflow. Workflow jobs are processed ahead of standard jobs because a user is usually waiting on the result.


Row Actions

The buttons at the right of each row depend on the state of the job.

Stop - Shown on a job that is queued or processing. Stops the job and marks it as an error with STOPPED BY and your user name in the status. A stopped job appears under Has Errors Only and can be re-queued later, alone or in bulk.


Start - Shown on a job that is still active but has reported an error. Puts the job back in the queue and OCRs only the pages that do not have text yet.


Reprocess - Shown on a finished job. Removes the existing OCR text for the document and re-processes every page. The document is not searchable until the new OCR completes.


Refresh - Shown on an active job. Refreshes that row immediately instead of waiting for the next automatic refresh.


Re-queueing Failed Jobs

When many jobs fail at once, for example after an outage, you can send them all back to the OCR queue in one step instead of restarting them one at a time. Only jobs that failed or were stopped can be re-queued this way. Jobs that are still queued or processing are never touched. If you need to re-run one of those, stop it first.

Checking jobs - Failed and stopped jobs show a check box at the left of the row. Click it to check the job, or Ctrl+click anywhere on the row (Cmd+click on a Mac). Shift+click a row to check every job between it and the last job you checked. The check box in the column header, and Check or Uncheck All Errored on This Page in the options menu, check every job on the current page that can be re-queued.


Re-queue Selected - Appears above the results once at least one job is checked. Re-queues the checked jobs.


Re-queue All Errored in This Search - In the options menu. Re-queues every failed or stopped job that matches the current search criteria, across all pages, not just the page you are looking at. Use it with a Submitted date range to target the jobs from one incident. The Not Complete Only and Not Started Only switches are ignored for this action because it always targets errored jobs.


Both actions open a confirmation that shows how many jobs will be re-queued and how many will be skipped, with the reason for each skip:

  • Still active - The job is queued or processing. Stop it first if you need to re-run it.
  • Completed without errors - Use the Reprocess button on the row instead.
  • Changed in the last 5 minutes - A job that was stopped or failed moments ago is skipped for one pass, in case its processing was still winding down. Run the action again after a few minutes.
  • No document on the OCR item - The document the job belonged to no longer exists.


Then choose how to re-queue:

  • Resume - The default. OCRs only the pages that do not have text yet, so a job that failed part way through picks up where it left off.
  • Full redo - Removes the existing OCR text for every selected document and re-processes every page. The documents are not searchable until their OCR completes again, so use this only when the existing text is wrong.


Two kinds of failure are marked by the system rather than by a person, and both can be re-queued the same way. A job whose queue request could not be handed to the processing pool after repeated attempts shows "Dispatch to Batch failed after 5 attempts". A job that support marked as lost shows "Marked failed by", the support person's name, the date, and the reason.


Re-queued jobs go to the back of the OCR queue and are processed at the normal rate for your site. Their rows change to Restarting and then Queued, and refresh as they progress. Up to 5,000 jobs are re-queued in one run; if more match, the confirmation says so and you can run the action again once those have started.