Albanian
Albanian handwriting, print and signage turned into text, whatever alphabet it was written in
Until the 1908 Congress of Manastir settled on a single Latin alphabet, Albanian had been written in Greek and Arabic letters and in at least three invented alphabets of its own, so a page from before that date can turn up in a script no other Albanian document uses. Modern Albanian solves the same problem the other way round, spelling sounds Latin never had with digraphs such as sh, zh, nj and rr and with the extra letters ë and ç. Evidano reads both ends of that range and sets the transcription beside the image so every unfamiliar letter or digraph can be checked.
Sample pages
Real pages from public collections, shown beside their transcriptions. Pages whose transcription is still being checked are marked.

Source · Public domain · Naum Veqilharxhi
Transcription being verified
Vithkuqi’s letter values are documented by specialists such as Robert Elsie, but no published line-by-line transcription of this specific title page has been located.
What the model is told to watch for: Produce a diplomatic transcription in Albanian.

Source · CC BY-SA 3.0 · AceDouble
Transcription being verified
The declaration’s Albanian text is quoted and paraphrased in many histories of Albanian independence, but an edited, line-matched transcription of this exact facsimile, including its Ottoman-script line, has not been located.
What the model is told to watch for: Produce a diplomatic transcription in Albanian.

Qemal Baholli Public Library, Elbasan, 23 March 1923 · Source · CC BY 4.0 · Buona1, Wikipedian in Residence
Transcription being verified
No independent transcription of this issue has been published; the library’s digitisation project describes but does not transcribe individual pages.
What the model is told to watch for: Produce a diplomatic transcription in Albanian.

Transcription being verified
A single business sign; no separate published transcription exists for it.
What the model is told to watch for: Produce a diplomatic transcription in Albanian.

Source · Public domain · Published by the Drita society; scanned by Google Books
Transcription
The title and all fourteen lines of verse on the page. The Wikisource text is a proofread transliteration into the modern Albanian alphabet rather than a diplomatic copy of the Istanbul-alphabet letterforms, and it carries the printed acute and grave accents; the line numbers Wikisource prints beside lines 5 and 10 are editorial, are not on the page itself, and are left out here.
BAGËTI E BUJQËSIJA O málet' e Shqipërisë! e ju ô lísat' e gjatë! Fúshat e gjera me lule! q' u kam ndër mënt dit' e natë, Ju bregore bukuroshe! e ju lumënjt' e kulluar! Çuka, kodra, brinja, gërxhe dhe pylle të gjelbëruar! Dotë këndónj bagëtinë që mbani ju e ushqeni, O vë́ndëthit' e bekuar! ju më́ndjenë ma dëfreni. Ti Shqipërí, më ep nderrë, më ep emërin Shqipëtár, Zë́mërnë ti mả gatove plot me dëshirë dhe me zjar. Shqipërí! ô mëma ime! ndonë se jam i mërguar, Dashurinë t' ënde kûrrë zë́mëra s' e ka haruar. Kur dëgjón zëthin e s' ëmës qysh e le qengji kopenë, Blegërín dy a tri herë edhe ikën e mer dhenë, Edhé në iprefshin udhën njëzét a tridhjetë vetë, E tả trëmbin, ajý s' kthehet, po shkon në mest si shigjetë,
Transcription from Bagëti e Bujqësija (Bucharest, 1886), page 3, proofread text via Wikisource (CC BY-SA 4.0).
What Albanian text looks like on the page
Centuries of Ottoman rule left Albanian without an official alphabet, so where it was written down at all before the twentieth century, a writer typically reached for whichever script their own religious and educational tradition used: Greek letters among Orthodox Albanians in the south, the Arabic script among Muslim Albanians and in dealings with Ottoman administration, and Latin letters, in a range of competing spelling systems, among Catholic Albanians in the north and the Italian-facing diaspora. Alongside these borrowed alphabets, Albanians also invented scripts built for the language itself: Elbasan, known from an anonymous eighteenth-century gospel manuscript; Todhri, devised in the early nineteenth century by Todhri Haxhifilipi of Elbasan; and Vithkuqi, designed in the 1840s by Naum Veqilharxhi specifically to give Albanian letters owing nothing to Greek, Turkish or Slavic tradition.
The Congress of Manastir, held in November 1908, ended the competition by adopting a Latin alphabet for Albanian, essentially the one still in use, and every one of the older scripts fell out of use within a generation. That alphabet solves the mismatch between Albanian’s sounds and the Latin letter set with nine two-letter digraphs — dh, gj, ll, nj, rr, sh, th, xh and zh — each standing for one sound and each alphabetised as a single letter, plus two additional letters of its own, ë for the central vowel and ç for a sound close to English ch.
Everyday modern material is printed or written in that standard orthography, but reading it still means recognising the digraphs and the two extra letters reliably: a worn photocopy, a strict ASCII keyboard or a foreign typesetter can drop ë and ç first, and a document from the transitional years around independence can still mix Albanian Latin script with Ottoman Turkish in Arabic script on the same page, a reminder that the 1908 standard took time to reach every kind of document.
Why ordinary OCR struggles here
A pre-1908 page could be in almost any script
Older Albanian material cannot be assumed to be in Latin letters at all: a document might be in Greek-lettered Albanian, Arabic-lettered Albanian, or one of the invented Elbasan, Todhri or Vithkuqi alphabets, none of which share letterforms with each other or with the modern alphabet.
Digraphs that are really one letter
Sh, zh, nj, rr, dh, gj, ll, th and xh each spell a single sound and are alphabetised as one letter, but a system reading letter by letter can split them at the wrong point or fail to treat them as the units they are.
Ë and ç easily lost or swapped
The two letters Albanian added to the Latin alphabet are often the first casualties of a worn scan, a keyboard without the right key, or a system that quietly falls back to plain e and c, and losing either one changes the spelling and often the meaning of a word.
Mixed scripts on one official page
A document from the transition years, such as the 1912 Vlorë declaration, can carry Albanian in Latin letters, a line of Ottoman Turkish in Arabic script, and signatures in either script side by side, so a transcription has to keep each block in its own script rather than merge them into one.
Who works with this material
Albanian genealogists and local historians
Church, school and land records survive in Greek-lettered, Arabic-lettered, Latin-lettered and invented-alphabet Albanian depending on the region and religion of the writer, so working out which alphabet a record uses is often the first step in reading it at all.
Historians of Albanian nationalism and the Rilindja
The movement to give Albanian a single national alphabet is documented in exactly the kind of primers, pamphlets and congress records this page draws its earliest sample from, and reading them means recognising scripts that fell out of use within a generation.
Archivists of early twentieth-century Albanian institutions
Newspapers, gazette notices and official correspondence from the years around independence often mix scripts and spelling conventions within a single archive, and cataloguing them accurately depends on transcribing each block in the script it was actually written in.
Diaspora and community organisations
Shop signs, notices and correspondence in Albanian communities abroad need ë, ç and the digraphs read correctly before they can be translated, indexed or searched.
Getting the best transcription
Identify the alphabet before transcribing
Say whether a page is in Latin, Greek, Arabic script, or one of the invented Elbasan, Todhri or Vithkuqi alphabets, since none of them share a letter set with any of the others.
Ask for digraphs to be kept as single units
Specify that dh, gj, ll, nj, rr, sh, th, xh and zh should be transcribed and, where relevant, alphabetised as the single letters they represent rather than as two separate ones.
Confirm ë and ç are preserved
Tell the model that ë and ç are distinct letters in modern Albanian, not accented variants that can be dropped or replaced with plain e and c when a font or keyboard makes that easier.
Flag documents that mix scripts
On a transitional document, ask for each script to be transcribed as its own block, so an Albanian Latin passage, a line of Ottoman Arabic script and a set of signatures are not run together as if they were one continuous text.
Frequently asked questions
- Can it read Albanian written in Greek or Arabic letters, or in an invented alphabet like Vithkuqi?
- Yes, when you say which script the page uses. Greek-lettered and Arabic-lettered Albanian, and the Elbasan, Todhri and Vithkuqi alphabets, are each read on their own terms rather than assumed to be the modern Latin alphabet.
- How does it handle Albanian’s digraphs?
- Dh, gj, ll, nj, rr, sh, th, xh and zh are recognised as single letters, each standing for one sound, rather than read and alphabetised as two separate letters.
- What happens with ë and ç?
- They are kept as the distinct letters they are rather than dropped or replaced with plain e and c, which matters because the difference changes both spelling and meaning.
- Can it transcribe a document that mixes scripts, like the 1912 Declaration of Independence?
- Yes. Albanian Latin script, Ottoman Turkish in Arabic script, and signatures in either script are transcribed as separate blocks in the order they appear, rather than merged into one passage.
- Does it know the difference between pre- and post-1908 Albanian spelling?
- Yes, when you flag the material as pre-standardisation. Older spelling conventions and script choices are read as period-correct rather than corrected to the alphabet fixed at the Congress of Manastir in 1908.
Related scripts and pages
Transcribe your Albanian documents
Upload a document, a newspaper page or a sign, in Latin, Greek, Arabic or an older Albanian alphabet, and check the transcription line by line before you export it.
