Can it tell the main text from a gloss written round the edge of it?

Yes, when the prompt asks for them separately. Each marginal note is returned as its own block with its position on the page, so a note running up the outer edge does not end up spliced between two lines of the work it comments on.

How reliable is it on names in an isnād?

Chains of transmission are the hardest part of any ḥadīth manuscript, because a name gives no context to correct a misplaced dot. Uncertain readings are marked rather than resolved silently, which lets you check the doubtful ones against a biographical dictionary.

Will vocalisation be reproduced where the scribe wrote it?

It can be. Many codices are only partly vocalised, and the words the scribe chose to point are often the ones he thought a reader would get wrong, so the pattern is worth keeping. You can also ask for a plain unvocalised text for indexing.

Are colophons and reading certificates handled differently from the text?

They are transcribed as separate records and can be returned with the names, places and dates picked out. A single manuscript often carries several certificates in different hands, and each is kept apart rather than run together.

What happens with a catchword at the bottom of the page?

It is recognised for what it is and labelled, not appended to the last line of text. Because it repeats the first word of the next page, keeping it also lets you check that a manuscript has been imaged in the right order.

Arabic · c. 1000–1500

Read the book hand, the margins and the notes at the end

A medieval Arabic codex rarely holds still. The main text sits in a ruled block in naskh; glosses circle it at three different angles, a corrector has squeezed a word above the line, and the last page carries a colophon, a reading certificate and two ownership notes in hands two centuries apart. Evidano transcribes each of those as its own text, in its own place on the page.

Upload a page

Sample pages

Page of a small Qurʾan in fully vocalised naskh with two gold sura headings and gold rosettes marking the verse ends — Sample 1. The Qurʾan copied and illuminated by ʿAlī ibn Hilāl, known as Ibn al-Bawwāb, at Baghdad in A.H. 391 (A.D. 1000–1001), Dublin, Chester Beatty Library, Is 1431, f. 279r. It is the earliest surviving Qurʾan in a cursive book hand, and the model for what naskh became: full vocalisation, gold rosettes at the verse ends, sura headings in a gold display script and a marginal medallion for every fifth and tenth verse.

Facsimile plate of two manuscript pages side by side, the left ending a text in a large, widely spaced hand above a block of small crowded lines, the right headed by a dense block with spaced-out names below it — Sample 2. Plate 165 of B. Moritz, Arabic Palaeography (Cairo, 1905), reproducing two pages of A.H. 504 (A.D. 1111) from the Khedivial Library, Cairo, which Moritz indexes under the name al-Mubārak ibn Aḥmad al-Anṣārī. On the left-hand page a large, widely spaced naskh closes the text and names the copyist, and a samāʿ certificate added afterwards in a much smaller and faster hand is squeezed into the foot; the page beside it carries a dense block of the same small writing at its head, with the names of those who heard the book read spaced out below.

Facsimile page whose text block carries lines of small annotation written at a slant across its head and down its outer margin — Sample 3. A page of al-Qalqashandī’s chancery encyclopaedia Ṣubḥ al-aʿshā, dated A.H. 799 (A.D. 1397), al-Azhar Library, from plate 171 of Moritz’s facsimile atlas. Four section headings in a large display hand, al-qāʿida al-ūlā and its successors, break up the block, technical terms are vocalised where the author explains how a place name is pronounced, and the margin is filled with annotation that turns the corner and runs on.

Facsimile of the closing page of a manuscript, with the final lines tapering to a point above two separate blocks of later notes — Sample 4. The last page of a copy of al-Ghazālī’s al-Basīṭ, dated A.H. 636 (A.D. 1239), Khedivial Library, Cairo, from plate 140 of Moritz’s atlas. The text ends in a tapering triangle with the formula tamma al-kitāb bi-ḥamdi llāh; below it the copyist’s dated note, and lower still an ownership or endowment note added later in a coarser, looser hand.

Illuminated Qurʾan page with six lines of large vocalised naskh between two gold and blue cartouches, gold rosettes between the verses — Sample 5. Sūrat al-Fātiḥa in Mamluk naskh, Egypt, fourteenth or fifteenth century, Washington, Library of Congress. The cartouches above and below give the sura’s title and its verse count in gold outline; six-petalled gold rosettes divide the verses, and the small red lā above the second line is a recitation mark telling the reader not to pause there. Three extra verse markers were written at the foot in error and the word āmīn added to patch the mistake.

What a Mashriqi codex page looks like

From the tenth century onwards the scripts used for books in the eastern Islamic world were codified as a set of proportioned hands, of which naskh became the ordinary book hand: a rounded, fully cursive script with clear positional forms, a firm baseline and a modest contrast between thick and thin. Around it sit its relatives — the taller, more elegant rayḥān, the heavy muḥaqqaq and thuluth used for headings and grand Qurʾans, and the small, quick taʿlīq of the scholar’s working copy. Pages are ruled with a misṭara and written in a block whose width is fixed; a catchword at the foot of the verso gives the first word of the next page so that the quires can be assembled in order.

Pointing is standard by this period, but it moves. Dots are placed high, low or squashed into the space above a neighbouring letter, and two dots often become a single stroke. Undotted letters may be marked as such with a tiny sublinear version of themselves or with a small caret, which is easy to mistake for a vowel. Vocalisation is partial: a Qurʾan or a poem is normally fully pointed, a legal or historical work is not, and a scribe will vocalise a rare word or a proper name in an otherwise bare paragraph. Above the line sit shadda, sukūn and the superscript alif; below it, the kasra. The lām-alif ligature and the tighter combinations that stack a letter above another change the order the strokes are made in, so the reading order of a compressed word is not the order the eye first sees.

The apparatus around the text matters as much as the text. Section headings are written in red, gold or a larger hand; the words qāla and faṣl are often rubricated; a triangle of dots or the word ṣaḥḥa marks a checked reading, and balagha shows how far a collation reached. Glosses run up the outer margin and turn the corner, and a long note may spiral. At the end come the colophon, giving the copyist, the place and the date; the samāʿ or ijāza certificates recording who heard the book read and when; and later ownership, endowment and purchase notes, each in a different hand and often a different ink.

Why ordinary OCR struggles here

Letters that differ only by their dots

Once a scribe writes quickly, the two dots of tāʾ merge into a dash, the three of thāʾ into a caret, and a dot drifts over the wrong tooth. On a page of unfamiliar proper names there is no vocabulary to fall back on, so the transcription has to be able to say that a skeleton was read as one letter rather than another and that the reading is not certain.

Marginalia at every angle

Glosses are written up the side of the block, along the top, diagonally into the corner and sometimes upside down. Ordinary layout analysis either drops them or splices them into the main text between two lines that have nothing to do with them. Each note has to be recognised as a separate text, with its position and its anchor recorded.

Ligatures that reorder the strokes

In compressed words a letter is written above its neighbour rather than beside it, and the lām-alif and similar combinations are drawn as a single shape. A recogniser working stroke by stroke reads such a cluster in the order it meets the ink, which is not the order of the letters, and produces a word that does not exist.

Everything that is not the work

A catchword at the foot of a page, a collation mark, a colophon, a reading certificate and a nineteenth-century librarian’s stamp all sit on the same folio as the text. Merged into one stream they corrupt the edition; separated and labelled they are some of the most useful evidence the manuscript carries.

Who works with this material

Editors of classical Arabic texts

Preparing a critical edition means collating several witnesses line by line. A transcription that keeps the manuscript’s own spelling, marks its corrections and reproduces the marginal variants gives an apparatus that can be built up rather than reconstructed from memory.

Historians of books and reading

Samāʿ and ijāza notes record who read a book, to whom, where and in what year. Read in bulk they map a teaching network across cities and generations, which is only possible if the notes are transcribed as separate texts with their dates and names intact.

Manuscript catalogers and digital libraries

Catalogue records live or die on the colophon. Pulling the copyist’s name, the place and the date out of the last page of every volume in a collection turns a shelf list into something a researcher can search.

Codicology and palaeography teaching

Students need to see that a page is layered. Showing the block, the glosses and the later notes as three separate transcriptions makes the structure of the page obvious in a way that a photograph alone does not.

Getting the best transcription

Step 1

Name the work and the discipline

Say that the page is fiqh, ḥadīth, lexicography or poetry. The technical vocabulary, the isnād formulae and the abbreviations differ sharply between them, and knowing which to expect settles many of the ambiguous skeletons before they become errors.

Step 2

Ask for the margins separately

Request the ruled block first, then each marginal note as its own text with a note of where it sits and which line it comments on. Interlinear glosses should be tied to the word above which they are written rather than dropped into the line.

Step 3

Set the spelling policy in advance

Decide whether final ة is to be transcribed as written or normalised to ه, whether ى without dots is to be resolved to ي, and whether hamza is to be supplied. Fixing this once keeps a whole codex consistent and makes the output searchable.

Step 4

Flag the furniture

Ask for catchwords, rubricated headings, collation marks, the colophon and any certificates or ownership notes to be labelled as such. They are what a cataloguer needs and what an editor must keep out of the text.

Frequently asked questions

Can it tell the main text from a gloss written round the edge of it?: Yes, when the prompt asks for them separately. Each marginal note is returned as its own block with its position on the page, so a note running up the outer edge does not end up spliced between two lines of the work it comments on.
How reliable is it on names in an isnād?: Chains of transmission are the hardest part of any ḥadīth manuscript, because a name gives no context to correct a misplaced dot. Uncertain readings are marked rather than resolved silently, which lets you check the doubtful ones against a biographical dictionary.
Will vocalisation be reproduced where the scribe wrote it?: It can be. Many codices are only partly vocalised, and the words the scribe chose to point are often the ones he thought a reader would get wrong, so the pattern is worth keeping. You can also ask for a plain unvocalised text for indexing.
Are colophons and reading certificates handled differently from the text?: They are transcribed as separate records and can be returned with the names, places and dates picked out. A single manuscript often carries several certificates in different hands, and each is kept apart rather than run together.
What happens with a catchword at the bottom of the page?: It is recognised for what it is and labelled, not appended to the last line of text. Because it repeats the first word of the next page, keeping it also lets you check that a manuscript has been imaged in the right order.

Related scripts and pages

Early Qurʾanic and Kufic Codices, c. 700–1000The angular book hands that naskh replaced, and the layered diacritics they carry.Ottoman-Era Arabic Naskh and Ruqʿah Documents, c. 1600–1950The same book hand three centuries later, beside the fast cursive of registers and notes.Early New Persian Kufic and Naskh Manuscripts, c. 900–1200The same alphabet and the same book hands carrying a different language, with four extra letters.All Arabic hands and document typesEvery Arabic domain from the first papyri to Ottoman-era registers and printed Bulaq pages.Image to TextSearch all supported languages, scripts and document types.

Transcribe your Arabic codices

Upload a folio, choose Mashriqi Naskh and Related Book Hands, and ask for the block, the margins and the colophon as separate texts.

Read the book hand, the margins and the notes at the end

Sample pages

Evidano Transcription

Evidano Transcription

Evidano Transcription

Evidano Transcription

Evidano Transcription

What a Mashriqi codex page looks like

Why ordinary OCR struggles here

Letters that differ only by their dots

Marginalia at every angle

Ligatures that reorder the strokes

Everything that is not the work

Who works with this material

Editors of classical Arabic texts

Historians of books and reading

Manuscript catalogers and digital libraries

Codicology and palaeography teaching

Getting the best transcription

Name the work and the discipline

Ask for the margins separately

Set the spelling policy in advance

Flag the furniture

Further reading on this hand

Frequently asked questions

Related scripts and pages

Transcribe your Arabic codices

Company

About

Preview Evidano

Account

Newsletter