What is here, what has not been checked, and six specific things a reader who knows Syriac could settle in an afternoon.
Everything on this site was read off public-domain plates by one person who reads Latin and Greek — not Syriac — working with a language model, against standard reference grammar. Where a paradigm could be collated against Nöldeke it has been, cell by cell, and the collation is recorded in the record with its disagreements. Where it could not, the reading stands as proposed and says so.
No Syriacist has ruled on any of it. The registry carries the seat and leaves it empty: expert = "". That is not a disclaimer at the foot of the page; it is the reason for the page.
The asks below are deliberately narrow. None of them is "would you like to review a website." Each is a specific question with a specific record behind it, most answerable by someone who knows the language in the time it takes to read this page. They are in order of what would help most.
The highest-value item by a distance, because it is not re-extraction — it is pointing.
Twenty-seven entries were read from Payne Smith's 1903 Compendious Syriac Dictionary to fill vocabulary gaps the lessons opened. For each one the consonantal skeleton, the part of speech and the English gloss were read directly off the plate and are reliable. The vocalisation was deliberately left out. Payne Smith's plates are pointed and the points are visible at the resolutions used, but this project has no control table for that typeface's pointing and no second reader, and the standing rule here is never to read diacritics by eye and call them data. So voc is empty in all twenty-seven, on purpose, and each record says so on its face.
A reader who knows Syriac would be adding vowel points to twenty-seven skeletons that are already correct. Examples: ܐܒܝܠܐ mourner, ܩܕܫ sanctify, ܬܗܘܡܐ the deep. The full shard is on the sources shelf.
Ten of the twenty-seven also carry an uncertainty flag, and one of those is a plain coverage hole rather than a hard question: ܫܒܩ forgive was cropped from sense f) onward, so senses a)–e) and the entry's opening line were never read. That one needs one more look at a page already fetched, not expertise.
In the noun-states survey, row I.b, a word glossed 'foot' is printed with the masculine plural pattern (־ܝܢ / ־ܝ / ־ܐ) where Nöldeke's general rule at §70 predicts the feminine endings for a feminine noun of that shape.
The obvious explanation — that we misread the plate — has been ruled out: the row was re-verified on a second, tighter crop and the masculine pattern is genuinely what Nestle prints. So the remaining question is lexical, and it is beyond us. Is the lexeme misidentified? Or does Nestle really print it that way, and if so is it his slip or a real usage? This is flagged in the record rather than quietly "corrected," because correcting it on the strength of a memorised paradigm is exactly the failure this project is trying not to commit.
Nestle's §31 states the noun-plus-possessive-suffix rule and follows it with seven worked example forms, unlabeled. Nöldeke gives the same rule at §145 with a fully labeled ten-item example set, so the seven were mapped onto his named set by shape and length correlation and tagged accordingly — and the record marks every one of those tags proposed, not proven.
They are plausible. They are not read glyph by glyph at a resolution that would settle them. Someone who can simply look at the seven forms and name the persons would close this in a minute.
Nestle's preposition list prints two distinct words before a single shared gloss 'behind, after' — which the first pass recorded as one word covering both senses, and which a check against Nöldeke caught. Re-cropped at 5×, the plate does show two words, and the record now carries both of them separately.
What is still open is the second word's exact spelling: at that type size it is visually very close to the first, and the distinction turns on a teth against a taw. Recorded as uncertain rather than picked.
The manuscript track — a pocket Gospel book written at Vienna in 1554, the copy from which the first book printed in Syriac was set — explains its pointing to learners in four short glosses: seyame, the dot above, the dot below, and the abbreviation mark. Between them they carry most of the grammar a reader of that unpointed hand has to recover, and they are the notes a beginner will actually read.
All four are marked proposed. Each states counts drawn from the manuscript itself — the dot above marking the participle 167 times against 79 for the dot below, seyame falling on 1,182 of 2,475 lines, 675 word-forms out of 7,035 appearing in more than one pointing. The counts we stand behind. The grammatical explanations wrapped around them are the part that wants a Syriacist's eye.
Nestle's chrestomathy runs past Genesis into the Vitae Prophetarum and the Historia inventionis. Sixty-six pages are captured as records; twenty-two — the pages the lessons are built from — are fully keyed and independently diffed against a second printed edition. The other forty-four are partially or wholly unkeyed.
That is a different problem from everything above, and a bigger one: it needs either a reader who can key continuous Serto at speed, or a Serto HTR model good enough to draft it for correction. It is listed here for completeness and because someone reading this page may know that a usable model already exists, which would be worth more to us than the pages themselves.
Stated plainly, because it is the honest version of all six items: this project would like a Syriacist to take the empty seat — not to audit everything, but to be the person a genuinely undecidable reading gets sent to, a few times a year.
An answer goes into the record it belongs to, with the reasoning, and the reading's status moves from proposed to adjudicated. Any lesson built on that record is rebuilt from it. Contributions are credited by name in the record and on this page unless you would rather they were not, and everything here — records, lessons, drills — stays free, without an account or a paywall, permanently.
Barnes 1914 — and no licence is claimed over their text. A transcription of a public-domain work creates no new copyright.
is its own work: the lesson prose, the drills, the glosses and notes, and the encoding as a compilation.
truth behind the hand track — Cod. Syr. 1, from the HTR Winter School Vienna — ships a LICENSE.md containing the full text of CC BY-SA 4.0 beside an htr-united.yml declaring CC BY 4.0. Both sites state BY-SA, which is the conservative reading and is true under either. It has not been settled with the source. If you know who at the Winter School can answer that, saying so would itself be a help.
Records, plate readings and disagreements are all public on the sources shelf; what the course itself is and how it was built is on the about page. Corrections and offers to wilson.pruitt@gmail.com.