Beyond OCR: Why Past Attempts Fell Short

Digitizing the millions of pages related to the JFK assassination isn't as simple as running a standard Optical Character Recognition (OCR) program. For over sixty years, countless individuals and groups have attempted this, but the vast majority of results range from barely passable to completely unusable for serious analysis. The unique nature of these historical documents presents significant hurdles:

Close-up example of a scanned document showing faded text and background noise.
Degraded Quality

1. Degraded Quality

Many documents are scans of poor-quality photocopies, carbon copies, or originals suffering from decades of aging. Faded ink, bleed-through, stains, creases, and low-resolution scanning create noise and broken characters that fundamentally confuse traditional OCR algorithms.

Document example showing stamps, margin notes, and mixed text/table layout.
Complex Layouts

2. Complex Layouts

Government documents rarely follow simple text formats. They contain letterheads, stamps, handwritten annotations in margins, signatures, tables, forms, and interspersed images or diagrams. Basic OCR struggles to differentiate these elements from the main body text, leading to jumbled or omitted information.

Document example with heavy black redaction marks obscuring text.
Markings & Redactions

3. Markings & Redactions

Underlines, circles, cross-outs, arrows, and, most notoriously, heavy redactions obscure the original text. OCR might interpret these markings *as* text or simply fail to read the characters beneath or around them, resulting in incomplete or nonsensical output.

Close-up of a document section containing cursive handwriting.
Handwriting

4. Handwriting

A significant portion of the archive contains handwritten notes, comments, or entire handwritten pages. Standard OCR is designed for printed text and performs extremely poorly on variable human handwriting, often rendering it as gibberish.

The Need for a New Paradigm

Because of these combined challenges, simply applying OCR yields data that is fundamentally unreliable for large-scale analysis or detailed research. It requires significant manual correction, which is unfeasible for millions of pages. The JFK-Project recognizes these limitations and employs a sophisticated, multi-stage AI pipeline designed specifically to overcome these hurdles and achieve unprecedented accuracy.