The Charleston Mercury
The casualty register
Every named person the paper reported killed, wounded, missing, captured or dead of disease, 1860–1865 — with the wound, the weapon, the treatment and the outcome, each traceable to the column it was printed in.
The register is empty, and this is the honest reason.
The database, the controlled vocabularies and the extraction pipeline are built and working. What is missing is the newspaper. Of the 2,192 dates between 1 January 1860 and 31 December 1865, a copy of the Mercury could be found in public hands for 2 of them — a single Daily of 19 March 1862 and a single Tri-Weekly of 14 May 1863, both from the Lincoln Financial Foundation Collection.
Neither can be read. Both were scanned at about 110 to 140 dots per inch across a sheet some twenty inches wide, which puts a line of body type at roughly three pixels tall. At that size the letters are not faint, they are absent: there is nothing in the file to recognise. The repository’s own text for the 1863 issue runs to 170 bytes for the whole paper. A fresh pass with Tesseract 5.5.3 at 400 dpi recovered the masthead and almost nothing else, and enlarging the scan fourfold changed nothing, because enlargement adds no information that was not captured.
The discovery pass was run over what text there is. It returned 22 passages containing casualty language and judged all but one of them to be OCR noise, which is the correct verdict and the reason no person was created from any of them. A name misread is a man invented, and this register would rather hold nobody than hold somebody who was never there.
What it is meant to answer
The point of the thing is not a list of names. It is to be able to ask the paper a question and get an answer with a citation on it: who was killed at Battery Wagner; which men were struck by fragments of shell rather than by bullets; who lost an arm and who survived losing it; which of the wounded were hit while serving a gun, or carrying the colours, or standing on a deck when a shot brought the splinters down; and, not least, which of the men the Mercury reported dead turned out to be alive a fortnight later.
Three distinctions are built into the structure rather than left to whoever fills it in. A person is not a casualty event, and a casualty event is not an injury: one man wounded twice in one afternoon is one person, one event and two injuries, and the totals for people and for wounds are therefore different numbers and are always reported as different numbers. What the paper printed is never overwritten by what later research established; the two sit in separate columns, so a man reported killed on the fifth and found alive on the twelfth keeps both facts. And a wound that killed a man is marked fatal, while the graze on his forearm is not, merely because he later died.