Description: While the physical collection holds much value to the Institution and related communities, the organization and presentation of the digitized version has been hampered by electronic document formatting and accessibility compliance issues. Typefaces, paper type, fading and other damage, due to the materials available when it was created and the conditions in which it has been stored since, contribute to the difficulties in successfully deriving machine-readable or compliant public-facing digital versions of the historical items. When an attempt to “read” a particularly inaccessible document by using the institutional AI assistant was immediately successful, the collection organizer tested this approach on other documents in hopes of a solution to bypass instances of failed Optical Character Recognition (OCR).
Conclusion: With the possibility of including an AI-created PDF of text with the scanned image of the original document, rendering this vast group of old documents as a digitized historical collection that can be accessed by researchers around the world would become easier and quicker. Accessibility compliance would be fulfilled by the AI-created PDF of the text. Concurrently, the AI assistant was also directed to consider the individual documents as pieces of a larger collection so that synopses, keyword lists, biographical and geographic names could be rendered out of the data while it was being analyzed for text transcription. Using images of the electronic documents and the character recognition instances (both failed and successful) from this historically interesting collection, this poster presentation will diagram the steps the collection organizer took to perform this task, analyze its successes and failures, and explore other means by which AI can support access and accessibility of digitized historical collections.