Is Rashi script a handwriting or a typeface?

Both, in sequence. The letterforms are those of the Sephardi semi-cursive book hand, which scribes used for commentaries long before printing; the fifteenth-century punchcutters in Italy copied that hand, and the resulting type kept the name of the commentary it was most often used for.

Can the commentary be linked to the right verse?

Yes, when the lemma is recognised. The opening words of each note are matched against the biblical column on the same page, so a comment can be returned with the chapter and verse it belongs to, and a note whose lemma cannot be matched is reported rather than assigned by guesswork.

Are the vowel points and cantillation marks kept?

On the pointed central text they can be, including the accents, which matter for anyone working on the masoretic tradition. The surrounding commentary is unpointed and is returned as such, so the two do not have to share a convention.

What about the masoretic notes in the margin?

They are treated as their own layer. These notes are highly abbreviated and full of letters used as numbers, so they are returned separately with their position, and are worth checking by hand more often than the running text is.

Does a censor’s signature at the foot of the page get transcribed?

It is reported as a separate note, usually in Latin script and often with a date. Those signatures are evidence of where a copy was in the sixteenth or seventeenth century, so they are worth capturing even though they are not part of the book.

Hebrew · c. 1200–1800

Base text and commentary on one page

Sephardi scribes used the shape of their letters to mark rank: the biblical or legal text in a large pointed square hand, the commentary around it in a smaller, rounder semi-cursive. Printers copied that arrangement, cutting the semi-cursive into the type now called Rashi, so a single page can carry three texts, two scripts and one set of vowel points. Evidano keeps the layers apart and reports which block belongs to which verse.

Upload a page See the samples

Sample pages

Real pages from public collections, shown beside their transcriptions. Pages whose transcription is still being checked are marked.

A column of Hebrew in a rounded square hand, interrupted by a gold panel of large letters, with an illuminated border of blue and gold foliage — Sample 1. A column from the end of the Book of Knowledge in an illuminated Mishneh Torah copied in Spain or southern France in the first half of the fourteenth century and illuminated later at Perugia; National Library of Israel, MS Heb. 4°1193. The text is a broad Sephardi square hand written in two ruled columns; a gold panel carries the display words פרק עשירי that begin the last chapter, a list of chapter headings is set out in short lines, and the ruled border of the column runs into painted foliage.

A printed page of small rounded Hebrew type without vowel points, set as a single justified block on thick paper — Sample 2. A page of Rashi’s commentary on Genesis from the edition printed at Rome about 1469–72, one of the earliest Hebrew books to survive and probably the first appearance of the type later named after him. The whole page is set in the semi-cursive face, with no vowel points; a raised colon separates the lemma from the comment, and the lemmata are not otherwise distinguished, so the reader has to recognise where each new note begins.

A printed page with a large pointed Hebrew column in the middle and dense unpointed columns of small type down both sides — Sample 3. A page of Genesis from the second Rabbinic Bible, edited by Yaakov ben Ḥayyim and printed by Daniel Bomberg at Venice in 1524–25. The vocalised and accented biblical text runs down the centre with Targum Onkelos in the column beside it; Rashi, headed רש״י, fills the band across the top and the right-hand column in the small semi-cursive type, and the masoretic notes are printed in the narrow strip between the biblical text and the Targum. Water staining has eaten into the upper corner.

A paper page of evenly spaced Hebrew lines in a rounded hand, every word marked with vowel points, with short centred lines between paragraphs — Sample 4. A page from a manual of the laws of ritual slaughter and inspection copied by Mosheh de Milhaud in 1780; Herbert D. Katz Center for Advanced Judaic Studies, University of Pennsylvania, CAJS Rar Ms 495. The scribe writes a rounded, fully vocalised hand of the Sephardi type, lays out each ruling as a paragraph closed by a short centred line, and uses the same letters for the Hebrew of the rulings and for the Aramaic phrases quoted inside them.

The Sephardi hand and the type made from it

The Sephardi script family developed in Muslim and Christian Spain, in Provence and across North Africa, and after 1492 it went with the exiles to Italy, Amsterdam, Salonica and the Ottoman lands. Compared with the Ashkenazi hand it is rounder and more even: the pen is held flatter, so the contrast between thick and thin is gentler, the letters sit in a broad rectangle, aleph is drawn as a slanting stroke with two short arms, and the tails of final kaf and nun curve rather than drop straight. Scribes worked in three registers — a formal square hand for Bibles and prayer books, a lighter semi-square or beinoni for study copies, and a joined cursive for deeds, accounts and letters — and used them together, so that on a page of the Pentateuch the verses stand in square letters and Onkelos and Rashi run round them in a smaller mode.

When Hebrew printing began in Italy in the late 1460s the first punchcutters copied that smaller mode rather than the square hand. The type they produced was used for the commentaries in the earliest Pentateuchs, and because Rashi’s was the commentary most often set in it the type has been called Rashi script ever since — it has nothing to do with Rashi himself, who died three and a half centuries earlier. It is an unpointed face with narrow, rounded letters: the head of ד ends flush while ר curves away, ה is open at the left, ב keeps a squared foot where כ is round, and the sequence of a lamed and an aleph is often cast as a single sort. The classic result is the Mikraot Gedolot page, where a pointed and accented biblical column stands in the middle with the Targum beside it and two or three commentaries in the small type filling the outer columns, the head and the foot.

Both manuscript and print use the same shorthand. A geresh marks a shortened word and gershayim an acronym, but many contractions carry neither; rabbinic names collapse into initials; letters serve as numbers for folios, dates and cross-references; and the divine name appears as a pair of yods. Vocalisation is applied to the biblical text and left off everything else on the same sheet, so a system has to switch between reading with and without points several times a page. Copies that passed through Italy often carry a censor’s signature in Latin script at the foot, and study copies collect marginal notes in a later cursive that ignores the ruling entirely.

Why ordinary OCR struggles here

Two scripts, one of them pointed

The centre of the page is large, vocalised and regular; the commentary round it is small, unpointed and abbreviated. A model set up for one of those does badly on the other, and a single pass over the whole sheet tends to average the two into something that fits neither.

Letterforms that converge in the small type

At commentary size ד and ר, ב and כ, ה and ח differ by a fraction of a millimetre of ink, and worn or over-inked impressions close the gaps. Without vowel points there is no second signal, so the reading depends on recognising the word.

Abbreviations, initials and numbers

Rabbinic prose is written in contractions: authorities appear as initials, stock phrases as two or three letters, references as letters used for numbers. Expanding them is an editorial decision, and one made silently in the middle of a transcription cannot be undone later.

Reading order that is not linear

On a Mikraot Gedolot page a commentary begins beside the verse it explains, runs to the foot of its column, continues at the head of the next and may wrap round the central block. Read as parallel columns it produces four texts, none of them continuous.

Who works with this material

Editors of rabbinic texts

Commentaries, codes and responsa are collated across manuscripts and early editions in which the same passage is abbreviated differently. An editor needs the contractions left as they stand, so that a variant is not created or hidden by an expansion.

Cataloguers of Hebrew incunabula and early printing

Early editions are identified by the exact wording of a colophon, the presence of a printer’s formula and the state of a few sorts. Getting those short passages into text lets copies be compared without moving the books.

Historians of Sephardi communities

Communal registers, contracts and letters from Spain, the Maghreb and the Ottoman cities are written in the cursive mode of the same family. Names, sums and dates are what a researcher needs first, and they are the parts most often abbreviated.

Libraries digitising rabbinic collections

A shelf of unindexed volumes becomes usable once each page yields its opening words and its running heads. That is enough to match a scan to a bibliographic record and to find where a tractate or a section begins.

Getting the best transcription

Step 1

Describe the page as a layout, not a text

Say that there is a pointed central column, a Targum beside it and commentaries in an unpointed small type around the edges, and ask for each as a separate block. That one sentence prevents most of the errors a whole-page pass would make.

Step 2

Ask for the commentary keyed to its lemma

Rashi and his successors quote two or three words of the verse and then comment. Request that the lemma be marked, so that each note can be attached to the verse it explains rather than to the line it happens to start on.

Step 3

Fix the abbreviation policy in advance

Decide whether contractions are kept, expanded, or expanded in brackets, and whether letters used as numbers are also given as figures. Repeat the choice in every prompt: it is the setting that most affects how usable a batch is later.

Step 4

Test on a heavily inked copy

Early editions vary from page to page; some impressions are grey and some are so heavy that the counters of the letters fill in. Run one of the worst pages first, and see whether the small type comes back with an honest note of illegibility rather than a smooth guess.

Frequently asked questions

Is Rashi script a handwriting or a typeface?: Both, in sequence. The letterforms are those of the Sephardi semi-cursive book hand, which scribes used for commentaries long before printing; the fifteenth-century punchcutters in Italy copied that hand, and the resulting type kept the name of the commentary it was most often used for.
Can the commentary be linked to the right verse?: Yes, when the lemma is recognised. The opening words of each note are matched against the biblical column on the same page, so a comment can be returned with the chapter and verse it belongs to, and a note whose lemma cannot be matched is reported rather than assigned by guesswork.
Are the vowel points and cantillation marks kept?: On the pointed central text they can be, including the accents, which matter for anyone working on the masoretic tradition. The surrounding commentary is unpointed and is returned as such, so the two do not have to share a convention.
What about the masoretic notes in the margin?: They are treated as their own layer. These notes are highly abbreviated and full of letters used as numbers, so they are returned separately with their position, and are worth checking by hand more often than the running text is.
Does a censor’s signature at the foot of the page get transcribed?: It is reported as a separate note, usually in Latin script and often with a date. Those signatures are evidence of where a copy was in the sixteenth or seventeenth century, so they are worth capturing even though they are not part of the book.

Related scripts and pages

Ashkenazi Hebrew and Yiddish Manuscripts, c. 1200–1800The other medieval script family, and the books written in it.Ladino Solitreo Correspondence and Records, c. 1600–1950What the Sephardi cursive became in the hands of Ottoman letter-writers.Cairo Geniza Hebrew, Aramaic, and Judeo-Arabic Documents, c. 900–1300Where these semi-cursive letterforms can first be watched taking shape.All Hebrew and Jewish documentary handsMasoretic codices, Geniza fragments, Ashkenazi and Sephardi books, and Ladino letters in Solitreo.Image to TextSearch all supported languages, scripts and document types.

Transcribe base text and commentary together

Upload a Sephardi manuscript page or an early printed leaf, choose Sephardi Hebrew Semi-Cursive and Rashi-Type Hands, and say how the page is laid out.

Start transcribing Create a free account