Do today’s latest “AI” models offer capabilities not possible with traditional OCR, for unlocking documents whose handwritten contexts were impenetrable with previous technologies? A decade ago, the Office of the Historian scanned its “Consular Cards file”, a collection of 6,500 handwritten index cards containing listings of officials at all U.S. diplomatic and consular posts from 1789-1960. A unique and foundational source for understanding the history of U.S. foreign relations, the utility of the scanned cards remained limited due to OCR’s inability to extract text from the ornate cursive handwriting on these cards. Experiments conducted this year with multimodal AI tools have produced breakthrough, if imperfect, results. The talk will demonstrate the methodology and results of these experiments and will offer tips and caveats for scholars considering such tools.
Tools & Technologies Transcription