Starting a Digital Archive Project for Small Town Obituaries
Richmond, Indiana has been around long enough that the older newspaper records are falling apart physically. The Richmond Telegraph and the Palladium-Item microfilm collections hold decades of obituary information, but microfilm degrades. Ink fades on yellowing pages. The best obituary notices from the 1950s and earlier exist only in a handful of deteriorating copies held by the Wayne County Historical Society and the Richmond Public Library's local history room. Digital archiving is how these records survive. It is straightforward to describe and significantly more complicated to execute properly.
Richmond Indiana Obituaries How Digital Archives Are Preserving Local History And Memory
The basic process involves scanning or photographing physical obituary pages, running OCR text extraction, indexing the data into a searchable database, and making it available through a public interface. Most small-town historical projects I have seen follow something close to this pipeline, though the details matter more than the general outline. One thing beginners consistently miss is that obituaries are not uniformly formatted. A 1962 notice might be two paragraphs typeset in a narrow column with a handwritten correction in the margin. A 1998 notice is a full-width modern layout with a photograph. Your OCR software will handle these differently, and the metadata strategy needs to account for that variance from the start. If you assume uniformity, your search results become unreliable within a month.
Scanning and Capture
The choice between flatbed scanning and overhead book scanning depends entirely on the condition of your source material. Flatbeds produce higher quality results but cannot accommodate bound volumes without damage. Overhead systems like the Zeutschel or even a well-set-up DSLR rig on a copy stand preserve binding integrity and are faster for intact bound years of a newspaper. For Richmond-area obituaries specifically, the Wayne County Public Library microfilm collection is the primary source. Microfilm scanning requires a dedicated microfilm scanner with appropriate resolution settings. I would recommend at least 400 DPI for text legibility, though many archivists push toward 600 DPI when budget allows. The tradeoff is file size. A single year of microfilm at 600 DPI can easily exceed 200 gigabytes. One specific edge case I encountered involved obituaries that had been clipped from newspapers and handwritten onto index cards. The standard OCR engine treated the handwriting as noise and produced nothing useful. The workaround was to run the images through a two-pass process: first a standard Tesseract pass to capture the typed portions, then a manual transcription pass for the handwritten sections using a tool like Transkribus, which is trained on historical handwriting. It added roughly four hours of labor to a batch that would have otherwise taken thirty minutes, but it was the difference between having the record and losing it entirely.
Get the Full Details
OCR and Text Extraction
Optical character recognition for historical newspaper text has improved dramatically over the last five years, but it is not reliable without post-processing. Newspaper fonts from the early to mid twentieth century include serif faces like Times New Roman, Monotype Corsiva, and various Gothic typefaces that modern OCR models were not primarily trained on. The OCR will produce recognizable output most of the time, but expect a 5 to 15 percent error rate depending on print quality and age of the paper. Name normalization is critical for obituaries. People appear under multiple name variants across different publications. A woman might be listed as "Margaret Ann Smith Johnson" in one obituary and "Peggy Johnson" in another, both referring to the same person who died in 1987. Building a basic authority file with variant name matching before you deploy your search interface prevents the most common user frustration. A counter-intuitive point about OCR quality: scanning at higher resolution does not linearly improve accuracy past a certain point. Beyond roughly 600 DPI, the additional pixels do not help the OCR engine distinguish characters. Instead they increase processing time and storage costs proportionally. For dense newspaper text, 400 to 600 DPI is the practical range. Going higher is a waste unless you have unusually degraded source material.
Database and Indexing Structure
A flat text file is not an archive. You need a relational structure that captures the fields people actually search for. The essential fields for an obituary collection are: full name of deceased, date of death, date of publication, source publication name, page or column location, relationship mentions (surviving family members), and the full OCR text as a searchable field. Additional fields that prove useful over time include burial location, funeral home, cause of death (when mentioned), military service, and organization memberships. Do not build all of these upfront. Start with the core set and add fields as user queries reveal what people are actually looking for. I have seen projects that over-engineered their schema with fifty fields on day one and then spent two years maintaining data that nobody ever searched against. For the search interface itself, a basic full-text search on the OCR text combined with structured filters for date range and publication name covers the vast majority of queries. Adding faceted search for surname begins to matter once the collection exceeds roughly five thousand records. Before that threshold, a simple keyword search is adequate and simpler to maintain.
Access and Preservation
Digitizing the content is only half the work. Long-term digital preservation requires attention to file formats, checksums, and redundancy. TIFF is the standard archival format for master images. PDF/A is appropriate for born-digital document preservation. JPEG and PNG are acceptable for access copies but should not serve as your preservation master. Checksums on every file allow you to detect bit rot before it becomes a problem. Running periodic integrity checks on your storage array should be part of the maintenance routine, not something you remember after a failure occurs. The Library of Congress Open Archival Information System (OAIS) reference model provides a useful framework for this, even if you do not adopt it in full. There are real limitations to this approach. OCR will always struggle with damaged text, faded ink, and marginalia. Handwritten corrections, stamps, and tear marks frequently appear in historical obituaries and are invisible to automated systems. The metadata you create is only as good as the source material, and many Wayne County records from the 1800s exist in fragments that no amount of scanning can fully recover.
A more comprehensive approach would involve partnering with the Indiana State Library's digital initiatives program or the Purdue University Digitization Services group. They have established workflows, institutional preservation agreements, and grant funding that small local projects cannot realistically replicate on their own. If your goal is simply to make the records findable, a standalone project can work. If your goal is long-term preservation, institutional partnership reduces the risk of the archive becoming another orphaned digital collection.
