factual
OCR Fails on Complex Historical Documents
Standard tools like Amazon Textract and Google Cloud Vision produce garbage on historical data because they are geared toward single-column books and cannot handle complex multi-column newspaper layouts with headlines, images, advertisements, and captions; without manual digitization, researchers only have keyword counts and cannot reconstruct sentences, so they cannot extract sentiment or how communities discussed events—a gap Dell is closing with computer vision and NLP methods to unlock massive historical datasets.
factualpending
Speaker
Melissa DellEvidence Quote
“we’ve been really investing in methods in computer vision as well as in natural language processing to process the output so that we can take data, historical data, on a large scale”
Created: 6/12/2026, 4:55:01 PM
Your notes
Loading notes...