Key Facts about Auto-Tagging
We strongly suggest reviewing guidance and results of generated content. Auto-tagging uses automated processes to generate document tags. While this functionality is designed to improve accessibility, the generated tags may not be fully accurate or complete in all cases. Instructors are responsible for reviewing and approving the generated tags before applying changes to the document.
Question | Answer |
|---|---|
Is this a third-party integration? | Yes. Ally leverages certain capabilities powered by a third-party SDK integration from Apryse to power the auto-tagging quick fix within the Instructor Feedback panel. (PDF tagging technology powered by Apryse™ and distributed by Blackboard under license. All rights reserved.) |
Is this an opt-in functionality? | Yes. Administrators need to activate the Auto-tagging capability in the LMS Ally configuration options. Administrators can activate or deactivate this functionality at any time. |
How is the system trained? | Blackboard is not involved in the training of any Apryse machine learning capabilities. These models are trained by Apryse. Ally does not further fine-tune any auto-tagging machine learning models using our own or our clients’ data. |
Is client data used for (re)training machine learning models? | No. Ally integrates this technology as an SDK and does not transmit client content externally to Apryse for processing or model training purposes. |
Is it possible to edit or approve tags individually after they are applied? | Currently there is no option to edit tags directly in the Instructor Feedback. Instructors can download the tagged file and edit outside of Ally leveraging Ally’s additional guidance if more improvements are needed. Similarly, there is no option to approve individual tags while leaving some things untagged. |
Does this auto-tagging feature replace the need for human review? | No, as with most accessibility issues, human review is still necessary and encouraged. Automation, like the technology used for automatic tagging, is often dependent on many factors and is not meant to fully replace human insight and review. Tags should be reviewed and verified as part of your accessibility procedures. This feature was designed to support accessibility efforts without replacing them or the humans involved. |
Does auto-tagging make a PDF fully accessible? | No, there are often additional improvements that can be made to the PDF, not only in terms of adjustments to tagging but also other issues that may be present in the document. For example, tagged PDFs with images now included may lack alternative text which could be added to continue to improve the accessibility for students. |
What could impact the quality of the auto-tags applied to my PDF? | There are a few factors that may impact the overall quality of tags applied. Depending on how your PDF was originally created, you may see a difference in structure and quality. Things like well-structured tables, lists, and headings can improve the result of the tagging process while the absence of this structure can result in some variation. In general, higher quality source files before conversion often result in better quality PDF tagging. Starting with an accessible source file before PDF conversion can help avoid many issues after the fact. |