Zentral- und Landesbibliothek Berlin

The Berlin central and regional library's digital collections, and for genealogy the thing that matters is THE BERLINER ADRESSBUCH, 1799 to 1943, complete, free, no account. A Berlin address book gives a householder's FORENAME, TRADE, POSTAL DISTRICT, STREET AND HOUSE NUMBER, year by year, so a household can be followed between censuses and a trade watched as it changes — which in the Berlin volumes of 1938 to 1943 is not a metaphor. Three sections per volume: the Namenteil (by surname), the Straßenteil (by street, listing a house's owner, manager and every tenant), and a trades section. THE POST-WAR BOOKS ARE HERE UNDER DIFFERENT IDENTIFIERS and are set in Antiqua, so their text layer simply works; the pre-1944 books are FRAKTUR and theirs does not. HOW TO USE IT WITHOUT AN ACCOUNT: a Solr query endpoint at /viewer/api/v1/index/query takes FULLTEXT: and PI_TOPSTRUCT: and returns page order numbers; /viewer/api/v1/records/<PI>/pages/<order>/text/ returns a page's OCR as IIIF annotations; and /viewer/content/<PI>/<width>/0/<8-digit>.jpg returns the page image at any width. The ALTO XML endpoint is permission-controlled and returns 403 for these volumes, so word coordinates are not available and a run must be found by cropping. FOUR THINGS MEASURED IN USE, ALL OF THEM ABOUT THE FRAKTUR TEXT LAYER. (1) SEARCHING FOR A RARE SURNAME IS NOT AN INSTRUMENT: the headword is routinely mangled past matching, and in one volume a street name came out as “Diessenbachsir”, so the year was found only by reading the run in place. (2) DIGITS ARE WORSE THAN LETTERS: a Fraktur 5 read as 6 and a 4 read as 1 were both caught only at the image, and a house number taken from the text layer should not be trusted. (3) THE LONG S is rendered as f or j, which is a help as well as a hindrance — it proves a street name that no spelling without an s could produce. (4) THE PER-PAGE OCR CAN INTERLEAVE TWO ADJACENT COLUMNS LINE BY LINE WITHOUT SAYING SO, silently attributing one surname's entries to its neighbour; the pages are four columns wide, a run can split across a column break with the headword repeated at the top of the next one, and an alphabetical check at the image is the only way to catch either. *** AND THE HOLDING IS SPLIT ACROSS FOUR RECORD IDENTIFIERS WITH A SIX-YEAR HOLE IN IT, WHICH IS THE FIRST THING TO KNOW: *** PI 34115495 is the Berliner Adreßbuch for 1903 TO 1943 and nothing earlier; PI 34115512 is the SAME TITLE for 1873 TO 1896; 1897 TO 1902 ARE NOT DIGITISED UNDER EITHER and return zero pages; PI 34117222 holds the post-war Namenteil and Branchen volumes of 1955 to 1970; and PI 15849358 the telephone books of 1970 to 1980. A query against the wrong anchor returns ZERO AND LOOKS EXACTLY LIKE AN ABSENT SURNAME, so run a common-surname control (Schmidt returns four figures a volume) before believing any zero. ONE MORE THING ABOUT THE PRE-1900 VOLUMES: THEY PRINT FORENAME INITIALS ONLY — “A., Beamter”, “O., Lederhändler”, “J., Ww.” — and a single letter is the worst-OCR'd thing on a Fraktur page, so a negative on a FORENAME cannot be drawn from them at all. Full forenames begin with the 1903 volume. *** AND THE RULE THAT TURNS A BERLIN ADDRESS INTO A REGISTRY OFFICE, MEASURED ON TWO HOUSES IN TWO INDEPENDENT SUBURBS. *** For a street in a suburb that ran its own administration, THE BERLIN STREET DIRECTORY GIVES THE COURT AND NOT THE STANDESAMT — "W. 62 Lutherstr. · A.G. Charlottb. — L.G. III. (geh. z. Charlottenburg)" — because the suburb kept its own offices. The answer is in the suburb's own part of the volume and it takes two pages: one list defines each Standesamt BY STADTBEZIRK, and a second maps every STREET AND HOUSE-NUMBER RANGE to a Bezirk. IN CHARLOTTENBURG A STREET NAME IS NOT AN ADDRESS — the HOUSE NUMBER selects the Bezirk, and the same street is split across several. Lutherstr. 12-43 is one Bezirk and the rest of the street is not; Joachimsthalerstr. runs 1-4, 5-9 and 10-21 across three. A reading that names the street without the number has not finished. AND THE TWO VOLUMES DIFFER IN WHAT THEY WILL ANSWER: the 1913 volume defines the offices by Stadtbezirk ("Standesamt I. Rankestr. 10. Umfassend die Stadtbezirke Hochschulviertel 1-7, Am Kurfürstendamm 1-6, Ostviertel 1-8"), while THE 1903 VOLUME NAMES ITS OFFICES AND THEIR REGISTRARS AND NOT THEIR DISTRICTS AT ALL — so a 1903 address can be resolved to a Bezirk but NOT from that volume alone to an office. CHARLOTTENBURG HAD TWO REGISTRY OFFICES IN 1903 (Rankestr. 10 and Kirchstr. 10) AND THREE BY 1913, so a mapping read off one year must not be carried to the other. Finally the numerals are the first thing the Fraktur OCR loses: a range printed 5-9 came through as "5 -'i" and a house number 34 as 31, so the figure that decides the office is exactly the figure that must be read at the image.