Why I Keep a Folder of Scanned Brockton Paper Back Issues
I have a box of photocopies from the Brockton Enterprise going back to 1987. Most of it is obituaries. I pulled them out last month because someone was researching a great-uncle who died in 1992 and the family had no record of what year he passed. The index card in my filing cabinet said 1991. The actual newspaper clipping said November 1992. This kind of gap is exactly why I keep doing this work. Local newspaper obituary archives are not automatic. They do not catalog themselves. When a funeral home sends a submission to the Enterprise, it runs through typesetting, runs past the family proofreading stage, and then gets printed on newsprint that yellows within three decades. The physical copy survives because someone put it in a drawer. Not because the newspaper has an archiving policy. The process I use starts with the microfilm. Brockton Public Library has the Enterprise on reel microfilm going back to the mid-1970s. I take photos with a DSLR on a copy stand, about 400 DPI, raw format. I then OCR each page with Tesseract, but I never trust the output without manual correction. Obituaries have names with accents, old-fashioned spellings, and religious affiliations that the OCR software routinely mangles. The name Maureen becomes Mauraean. Fitzpatrick becomes Fitzpatrick. I correct these in a spreadsheet before building any database.
Here is the part nobody tells you: the real problem is not digitization. It is the metadata gap. An obituary page contains the deceased name, the date of death, the funeral home, the surviving family list, and the service location. But when I try to cross-reference a name across multiple years, the OCR errors compound. Someone named Josephine Marini might appear as Josefine Marine in one issue and Josephine Marino in another. I built a fuzzy matching script using Levenshtein distance with a threshold of 2 edits, but even that misses cases where the first name itself is misspelled. The workaround I settled on is building a confidence-weighted index where each variant gets linked to the most likely canonical name based on family member matches across adjacent obituaries. I ran into a specific edge case last spring. A researcher asked about a Thomas P. O'Brien who died around 1978. The microfilm was scratched in the death notices section for March through May of that year. Three months of obituaries gone. I called the Enterprise archives department and found they had a separate file box of carbon-copy submissions that predate the microfilm. The scratch was only on the library's copy. The newspaper's own archive had the complete run. This is not a universal problem but it happens often enough that I always check multiple sources before declaring a gap permanent. The indexing system I use now stores each obituary as a JSON file with fields for full name, death date, birth date (when available), funeral home, cemetery, surviving relatives, and source citation. I keep the raw OCR text alongside the corrected version so anyone can audit my work. The database runs on SQLite with full-text search enabled on the name and relative fields. Querying for all O'Briens who died between 1970 and 1985 returns results in under 200 milliseconds on a standard laptop.
There are limitations worth stating bluntly. Newsprint obituaries from the 1960s and earlier are often on low-quality paper that degrades regardless of storage conditions. Ink fades. The microfilm does not help when the original was already illegible. I have spent hours trying to read a name that simply cannot be recovered. In these cases, I note the gap in the index rather than guessing. The alternative is a flawed record that looks complete but is wrong, which is worse than having nothing at all. Another bottleneck is the funeral home submission pipeline. Not every family publishes an obituary. Some arrange private services with no public notice. The archive captures only what was submitted, which skews the demographic picture toward families who could afford the per-line cost and chose public notice. I do not claim completeness. I claim verifiable coverage for what was printed. If you want to start your own archive, begin with the microfilm at your local library. Call ahead and confirm they have the run you need. Budget two weeks for the initial photo pass if you are doing it manually. The scanning equipment costs about $400 for a decent copy stand and ring light if you already own a camera. Factor in roughly 4 hours per month for ongoing OCR correction and indexing if you are adding new issues weekly.
Get the Full Details
The entry point is simple: pick a neighborhood, pull the relevant years, and start correcting names. The work compounds. Each corrected obituary makes the next one easier because family member cross-references build up over time. Five years in and you have a searchable network of related deaths that would take days to reconstruct from scratch.
