Classical Near East Texts
Explore, search, and analyze the literary, documentary, and epigraphic corpora of the Classical Near East across languages, cultures, and religious traditions.
Prototyped on the Syriac homilies attributed to Isaac of Antioch, and already reaching beyond them to other corpora.
What the tool is for
The tool is meant to give a comprehensive jumpstart on any research question for any corpus whose data has been loaded into it. The tool's most advanced feature is the Cross-Corpora Results page. For each question a researcher asks, it gives accurate, interrelated AI summaries of the data, and an organized layout of the relevant hits, line by line. The researcher can then use these as an aid in examining the primary sources more closely.
The tool has three parts, linked at the top of this page and in the bar above every page of it. It is being built first on one corpus, the homilies attributed to Isaac of Antioch, which is why much of what follows is about Isaac; the same methods are already being applied to other corpora. Every page has an About this page box at the top: open it for what the page shows and how to read it.
Why this is more reliable than asking a general AI chatbot
A chatbot such as ChatGPT answers from what its model absorbed in training (trillions of words), perhaps topped up by a web search or the few files a user can upload into one chat. Its answer is a probable-sounding reply, with much room for error.
This tool reads the whole text of each corpus directly, every line. For Isaac, it reads the Syriac together with an English translation made by a multi-stage AI translation workflow: the project's own, or Jimmy Issa's AI translation (GPT 5.6 Sol), the two having been developed in collaboration. Every claim in a summary is tied to the hits it rests on, each cited the way its own corpus cites (for Isaac, the Bedjan page, the line, and a link to the printed scan), so a reader can check it.
Work in progress
The tool is honest about its limits. The summaries are written by AI and are provisional, and the hits are only as good as the OCR texts and word lists behind them. Work is under way on both. The OCR texts of the Syriac homilies, already mostly accurate, are being corrected line by line against the printed page; the AI translations are based on them. The matching is being improved so that it catches every true hit and no false ones, and a reader is given all the relevant passages and no irrelevant ones. These improvements will also make statistics about the corpus more accurate.
The prototype corpus: Isaac of Antioch
The prototype began with one research question of one corpus: where, and how, the Syriac homilies attributed to Isaac of Antioch argue against pagan religion and idolatry. Paul Bedjan's Homiliae S. Isaaci, Syri Antiocheni (Paris and Leipzig, 1903) prints homilies (memrē) under the name of Isaac of Antioch. The collection is not the work of one man. Following Adam Becker,1 the tool distinguishes four authors who came to share the name, from the early fifth century to the early sixth, and each homily page says which of them a homily is assigned to, where an assignment has been made.
The prototype covers the 67 memrē that Bedjan prints. For each, it gives the Syriac text, line by line, linked to the page of Bedjan's edition it comes from, with a link to the scan of that page on archive.org. It marks every occurrence of a curated list of 636 Syriac words and phrases that belong to the vocabulary of polemic: idols and idolatry, magic and divination, demons, defilement, deceit, and related themes.
Vision for the tool's expansion
The goal is to expand beyond the Isaac of Antioch corpus to late antique Syriac literature more broadly, and beyond that, to corpora of the other languages of the Classical Near East as well. To this end, the data of the Digital Syriac Corpus and of OCIANA (the Online Corpus of the Inscriptions of Ancient North Arabia) are already incorporated, and work is ongoing to incorporate data from other databases.
How to use it
The first four steps are for the Isaac of Antioch Corpus pages, the prototype; the last is for the Cross-Corpora Results, which cover every corpus.
Find the polemic in one homily. Open Isaac of Antioch Corpus and choose a homily from the index or the “Go to homily” menu. The homily page shows the Syriac text, page by page, with every matched word highlighted in its category's colour. The counts at the top say how many hits each category has and how dense they are per printed page. Pick a category from the menu to see only its hits.
See which homilies concentrate the polemic. Corpus Charts compares all homilies by hits or by density (hits per Bedjan page), for one category or all. The Heatmap shows every homily against every category at once, with similar homilies clustered together. From either page, Apply Current Filters to Corpus Word Search Page carries your selection over.
Search the whole corpus. Corpus Word Search finds a Syriac word or an English definition across all homilies, filtered by category and by homily.
Go to the printed page. Every line carries its Bedjan page and line number. View scan opens that page of Bedjan's edition on archive.org, so any reading can be checked against the print.
Read a research question across corpora. Open Cross-Corpora Results. Each row is one saved question: what it asks, which corpora it searched, how many documents and hits it found, and when it was run. Open a result to read every hit in context, filter it, or download it as a spreadsheet (CSV).
How to read what you see
Most of these notes describe the Isaac of Antioch Corpus pages; a note that also applies to the Cross-Corpora Results says so.
Homily numbers. Every homily has two numbers. Homily or B is Bedjan's number, in his printed order, and it orders everything in the tool. Memrā or m is the number used by syri.ac, which follows the checklist of Mathews, Jr. A homily is written “Homily 48 (Bedjan) · Memrā 131 (syri.ac / Mathews)”, or “B48 · m131” for short.
Search term categories. The curated words fall into the categories S1 to S15, plus S99 for words not yet placed:
S1 Paganism / Idolatry / Idols · S2 Magic / Divination · S3 Doctrinal Condemnation · S4 Satan / Demonic · S5 Defiling / Uncleanness / Impurity · S6 Deception / Deceit / Lying / Error / Falsehood · S7 Shame / Disgrace / Contempt · S8 Rebellion / Disobedience / Insolence / Transgression · S9 Oppression / Abuse / Harm / Injustice · S10 Sin / Sinfulness · S11 Wickedness / Evil / Impiety · S12 Corruption / Destruction · S13 Rejection / Repudiation · S14 Nazarene · S15 Jews · S99 Uncategorized Words
Flagged words. Where a word's definition contains a note in {braces}, the same spelling is also a different word (a homonym) that is not the one searched for; a note in [brackets] gives an alternate meaning. A coloured label beside a hit, such as “spelled like another word” or “other meanings too”, means “look twice”; hover over it to see why. Read every hit on these words in context before counting it.
Page and line numbers. A line is labelled like Memra Ln. 37 (p.589, Ln. 12): first the line's number within its homily (memrā), counted from the homily's first line, then the printed page of Bedjan's edition and the line on that page, counted from the top.
Density is a category's hits divided by the number of Bedjan pages in the homily, so long and short homilies can be compared.
Counts are matches, not readings. A hit means the word is there; whether the passage is polemical is a judgment the reader makes. Hits come from matching the forms of each word, and some forms are shared by unrelated words (for example, a form of ܢܟܠ deceit looks like ܟܠ all). Treat totals as a guide to where to read, not as findings. The homily pages and the Cross-Corpora Results count the same way (each word's forms, exact matches only), so their totals agree. On the homily pages a hit found under a different dictionary word is labelled, for example “under ܟܠ: check in context”.
AI-assisted readings. Some Cross-Corpora results search passages by what they argue (for example “idols and cult images” or “fate denied”), not by which words they use. These readings were made by an AI model and are provisional and not yet reviewed.
Where the translations and texts come from
Every translation names who or what produced it. On the Isaac pages, the English shown beside the Syriac comes from two sources, and every page that shows English says which:
- Homily 48 · Memrā 131: this project's own translation.
- Every other homily: Jimmy Issa's AI translation (GPT 5.6 Sol), from his complete AI translation of the homilies of Isaac of Antioch (compiled 2026-09-24). It is shown beside the Syriac, line by line, while the project's own translation of each homily is being prepared; this project's translation will appear in its place as each homily is finished.
The Syriac text of the Isaac homilies is an OCR reading of Bedjan's edition made by Jimmy Issa. He photographed the 837 pages of the edition, cleaned and cropped the images, and read them with the Qoruyo Syriac OCR models (its printed-page layout model and its Eastern Syriac text-recognition model) in the Kraken OCR engine. It is not the text that archive.org offers with its scan of the edition, which was made by general-purpose OCR (Tesseract) and is full of errors. The OCR as delivered was not corrected by hand; the project's own correction, line by line, is the work in progress described above. View scan on any line opens archive.org's scan of that printed page, so a reading can be checked against the print.
The Translation Workbench shows the project's translation work in progress, stage by stage. In the shared copy it can be read but not edited. What you see there is its state on 2026-11-22.
Other corpora come under their own terms:
| Corpus | Documents | Terms |
|---|---|---|
| Isaac of Antioch (Bedjan 1903) | 67 | Public domain |
| Digital Syriac Corpus | 632 | CC BY 4.0 |
| OCIANA (Safaitic and other Ancient North Arabian inscriptions) | 40,855 | Used with the permission of OCIANA; © The Ohio State University |
Summaries and readings on the Cross-Corpora pages are written by an AI model (Claude) during the project's reading pass. They are provisional and are being reviewed. Each one is marked as a summary, not a translation.
How to cite this tool
Justin Michael Wall. Classical Near East Texts: A Digital Platform for Comparative Textual Research, build of 2026-11-22, index of 2026-11-21. https://classicalneareast.org/. Isaac of Antioch corpus based on P. Bedjan, ed., Homiliae S. Isaaci, Syri Antiocheni (Paris and Leipzig, 1903); Syriac text read by Jimmy Issa with the Qoruyo Syriac OCR models in Kraken.
When you cite a result, cite the passage the way its own corpus cites it (for Isaac, by Bedjan page and line), and give the build date shown on the result page: the corpora are updated, and a later build may give a different count.
Notes
- Adam H. Becker, Isaac of Antioch: Homilies on Moral and Monastic Reform. Introduction, Text, and Translation, Writings from the Greco-Roman World 49 (Atlanta: SBL Press, 2024), 75. ↩︎

