The archive
What was left alone
The more interesting half of the work: which odd readings are the author’s own, and how they were told apart from the scanner’s.
The brief was to fix scanning artifacts and leave the author's wording, spelling and punctuation untouched. That distinction is the most consequential judgment in the whole job, so this file records where the line was drawn and why.
How the two were told apart
OCR errors and typing errors fail in different ways, and the difference is usually legible:
Scanning errors are visual. Tesseract confuses letters that look alike at 300 dpi on a worn typewriter ribbon — n/h, l/1, i/l, v/y, c/e, rn/m, q/g. So tne for the, nad for had, inguiry for inquiry, 3lst for 31st. These produce strings no typist would ever produce, and they appear once or twice rather than consistently.
Typing errors are phonetic or mechanical — a doubled letter, a dropped letter, a transposition, a misremembered spelling. So occured, embarassment, accomodated. These recur, because a writer who spells a word one way tends to keep spelling it that way.
Where the two tests disagreed, or where a reading could plausibly be either, the page image was opened and the question settled by looking. Anything still ambiguous after that was left as scanned, on the principle that an unwanted authorial quirk is recoverable later but an invented word is not.
Consistent authorial spellings
Each of these appears in the typescript as written, most of them more than once:
| Spelling | Times | Standard form |
|---|---|---|
occured |
7 | occurred |
socalled |
3 | so-called |
longitudenal |
4 | longitudinal |
anonimity |
3 | anonymity |
darkeness |
3 | darkness |
embarassment |
3 | embarrassment |
dilemna |
2 | dilemma |
disasterous |
2 | disastrous |
reconaissance |
2 | reconnaissance |
selfassurance |
2 | self-assurance |
accomodated, accomodations |
1 each | accommodated / accommodations |
aeronatical |
1 | aeronautical |
aggrevated |
1 | aggravated |
appearence |
1 | appearance |
assesment |
1 | assessment |
assidiously |
1 | assiduously |
battallions |
1 | battalions |
boyancy |
1 | buoyancy |
brakish |
1 | brackish |
capablities |
1 | capabilities |
caried |
1 | carried |
combattants |
1 | combatants |
concesssion |
1 | concession |
condtions |
1 | conditions |
consitute |
1 | constitute |
envisionaged |
1 | envisaged |
Penninsula |
1 | Peninsula |
propellors |
4 | propellers |
Portugese |
12 | Portuguese |
One-off typing slips, kept
Verified against the page image in each case — the typescript reads this way:
| Page (scan) | Reading | Standard form |
|---|---|---|
| 25 | mililtary airships |
military |
| 150 | 43 degress F. |
degrees — and degrees is spelled correctly on the very next line |
| 155 | broke through the clounds |
clouds |
| 155 | and alkso from the altimetric record |
also |
| 187 | the Bulgartian recruits |
Bulgarian — spelled correctly in the same sentence |
| 146 | allowing for necessary parations |
preparations |
| 253 | seems to have been into circulation |
put is simply missing |
| 166 | Wichester grand jury |
Winchester |
| — | Dartmout and others |
Dartmouth |
| — | Shorly after May 17 |
Shortly |
| — | snow-capped Kililmanjaro |
Kilimanjaro |
| — | Mirable dictu |
Mirabile dictu |
| — | the House of Hohenzoller |
Hohenzollern |
| — | An Imperial Commenddation |
Commendation |
| — | Straser sent a telegraph |
Strasser — correct in dozens of other places |
| — | Feburary 1919 |
February |
| — | Admiral Tripitz |
Tirpitz |
| — | a station at Kamina in Germanland |
German land |
| — | for the Mediterraenean area |
Mediterranean |
| — | Commander Mangelsdorff |
Mangelsdorf — the author uses both |
| — | the airship portection |
protection |
| — | gradully tiring, a surpise attack, it could be troubleseome |
gradually / surprise / troublesome |
| — | the sailors weren restless |
were |
| — | interrested in whether |
interested |
| — | the avalable information, radion triangulation |
available / radio |
| — | explosion of gas, ammuntion and fuel |
ammunition |
| — | as Sir Alfred had descibed it |
described |
| — | To guard minsweepers |
minesweepers |
| — | had probably suceeded |
succeeded |
| — | Weather and cicumstances |
circumstances |
| — | lay wainting in its hangar |
waiting |
| — | whatever suppplies it contained |
supplies |
| — | ropute provided for in the original plan |
route |
| — | With alighter ship |
a lighter |
| — | enough metereological information |
meteorological — he also writes meterological |
| — | a factually corect account of the voyage |
correct |
| — | may have acounted for this |
accounted |
| — | unable to develope them |
develop |
| — | firsat at the destruction |
first |
| — | across terrestial obstacles |
terrestrial |
| — | parabaloid shaped |
paraboloid |
That list is longer than the corrections log's letter-confusion section, and deliberately so. It is the clearest evidence that the transcription is faithful: after the scanning errors were removed, what remained was a writer's own typing.
British spellings
Left as typed rather than Americanised: criticised, emphasise, cyphers, and judgement — though the author is inconsistent here, using judgment 14 times against judgement 3 times. Both forms are his, so both stand.
The author's compound nouns
He habitually closes up compounds that most writers hyphenate or space. These are a stylistic habit, not errors, and they are consistent:
airforce · airforces · airdefense · airwar · airmasses · airlayers · airlegs · airstrength · airshipmen · groundcrew · bombload · machinegun · machineguns · rudderman · wingracks · teaparty · duraluminum
German and French left in place
von · der · des · Jetzt erst recht! · Der Feind hört mit! · vertretbare Schuld · Faust Natur · Hochwohlgeboren · Spähkorb · Reichsmarineamt · Admiralstab · Aktiengesellschaft · Afrika zu unseren Füssen · Zeppelin flieg. · Hilf uns im Krieg. · Quai d'Orsay · Deuxième Bureau · Pour le Mérite · lingua franca · askaris · elan
Umlauts were restored in these, because the scanner had mangled the diacritic itself into &, é, S or 8 — that is a scanning failure, not a spelling choice. See section 1 of corrections.md.
What remains uncorrected
About 0.46% of words in the finished document are not in a standard English dictionary. A large share is the material above — authorial spellings, German, and proper nouns from the manuscript's cast of several hundred people, ships and places. The remainder is a thin tail of single-occurrence OCR slips on faint lines that were judged not worth the risk of a guess. If you want to work through them, pipeline/tools/review_residue.py regenerates the review sheets that pair each remaining oddity with a crop of the line it came from.