The Buck Ends Here
My grandfather, Ivan D.E. Beaumont, wrote a column for the Nebraska City News-Press for about twenty years. I wanted every word of it back in readable text.
Why I built it
Grandpa ran and wrote for the Nebraska City paper for decades. His column was scattered across scanned pages in the county's digital archive: technically preserved, but practically unreadable. I wanted the family, and anyone else, to be able to sit down and actually read him again.
[Add your own line here, about what his writing means to you.]
What was hard
Standard OCR fell apart on decades-old newsprint: narrow columns, bleed-through, faded type, broken hyphenation. The output wasn't clean enough to read, let alone search.
Switching from plain OCR to an AI transcription step changed the quality completely. After that, the remaining work was a cleanup pass for what it still missed, and finding which piece on each page was actually his.
How it works
- 01CollectPull page PDFs from the county's digital archive.
- 02Find the column[How you located his column on each page.]
- 03Transcribe[Name of the transcription endpoint] instead of plain OCR.
- 04Clean & publishFix leftovers, date each piece, and publish the archive.
What I think now
This started as a family project, but the bigger lesson is that a huge amount of local history is stuck in scans nobody can search. Reading and cleaning old documents is now cheap enough to do at archive scale, whether for a county paper, a school, or a company's old records.