jayfurzy/orthodox-patristic-corpus

Orthodox Patristic Corpus

Released on the Feast of the Triumph of Orthodoxy, First Sunday of Great Lent, 2026.

---

Dataset Summary

The Orthodox Patristic Corpus is a 116M-token pre-training corpus of Orthodox Christian theological literature, assembled from the writings of 123 Church Fathers and Orthodox theologians spanning the 1st through 20th centuries. The corpus is primarily in Russian, drawing on the Azbyka.ru Orthodox digital library and other public-domain sources, with English and Greek patristic excerpts included.

The dataset was used to train Paterikon-3B, a domain-adapted language model for Orthodox Christian theology.

Total documents	67,046 (63,578 train / 3,468 val)
Total tokens	~116 million
Total size	~832 MB
Primary language	Russian (~98%)
Additional languages	English (~2%), Greek/Latin (patristic citations)
Authors	123 Church Fathers and Orthodox theologians
License	CC BY 4.0
Source	Azbyka.ru, CCEL, curated Orthodox text collections

---

Dataset Structure

Splits

Split	Documents	Approx. Tokens	File
train	63,578	~113M	cpt_corpus_train.jsonl
validation	3,468	~3M	cpt_corpus_val.jsonl
Total	67,046	~116M

Data Format

Each record is a JSON line with a single text field. The text begins with a structured header identifying the author, era, and language, followed by the theological passage:

{
  "text": "### St. Nicholas of Serbia (20th_century) [ru]\n\nК нему, устно и письменно, обращались за духовным советом весьма многие: священники и монахи, купцы и ремесленники..."
}

Header format: ### {Author Name} ({century/era}) [{language_code}]

Loading the Dataset

from datasets import load_dataset

# Full dataset
ds = load_dataset("jayfurzy/orthodox-patristic-corpus")

# Training split only
train_ds = load_dataset("jayfurzy/orthodox-patristic-corpus", split="train")

# Iterate
for example in train_ds:
    print(example["text"][:200])
    break

---

Corpus Contents

Sources

The corpus was assembled from three main source categories:

1. By-Principle Passages (~786,000 passages) Patristic texts organized by theological principle and topic, drawn from 123 authors. Each passage is a focused excerpt of 128–2,048 tokens on a specific theological theme.

2. Full-Length Works (7 texts, ~3M tokens) Complete patristic works included in their entirety, including major texts from the Russian Orthodox hesychast tradition and the Optina Elder corpus.

3. Thematic Corpora (55 collections, ~2M tokens) Curated topical collections on specific theological subjects, saints, or traditions.

Authors and Traditions Represented

Ancient and Byzantine Fathers (1st–14th century) St. John Chrysostom · St. Basil the Great · St. Gregory the Theologian (Nazianzen) · St. Gregory of Nyssa · St. Athanasius the Great · St. Cyril of Alexandria · St. John of Damascus · St. Maximus the Confessor · St. Symeon the New Theologian · St. Gregory Palamas · St. Macarius the Great · St. Ephrem the Syrian · St. John Climacus (Ladder of Divine Ascent) · St. Isaac the Syrian · Evagrius Ponticus · St. Dorotheus of Gaza · St. Abba Isaiah · and others

Russian Orthodox Fathers (18th–20th century) St. Theophan the Recluse · St. Ignatius Brianchaninov · St. Seraphim of Sarov · The Optina Elders (Leonid, Macarius, Ambrose, Hilarion, Anatoly I, Joseph, Barsanuphius, Anatoly II, Nektary, Nikon, Isaac II) · St. Paisios Velichkovsky · Metropolitan Anthony Khrapovitsky · Archbishop Averky Taushev · St. Nicholas of Serbia (Velimirović) · St. Silouan the Athonite · St. Paisios the Athonite · Bishop Nikolai Alexandrov · and others

Hagiographic Literature Lives of Saints (Menologion) by St. Dmitry of Rostov — comprehensive biographical accounts of Orthodox saints covering the entire liturgical calendar.

Theological Commentaries and Homilies Liturgical commentaries, homiletic collections, catechetical texts, and theological treatises on the sacraments, hesychasm, theosis, prayer, fasting, and Orthodox anthropology.

Language Distribution

Language	Documents	%	Notes
Russian	~62,253	~98%	Synodal-era and contemporary Orthodox Russian
Other (EN/EL/LA)	~1,325	~2%	CCEL translations, Greek patristic excerpts

The Russian texts span multiple registers: Synodal Russian (pre-1917), Soviet-era émigré theology, and contemporary Orthodox Russian. Church Slavonic elements appear in liturgical quotations within theological texts.

---

Dataset Creation

Collection Methodology

Texts were collected from:

1. Azbyka.ru — Russia's largest Orthodox digital library, hosting public-domain patristic texts, lives of saints, theological commentaries, and contemporary Orthodox writings. Crawled via structured web extraction with respect for rate limits.

2. Christian Classics Ethereal Library (CCEL) — English-language translations of Church Fathers; public-domain texts included.

3. Curated manual collections — specific authors and thematic corpora compiled from multiple Orthodox sources.

Preprocessing

1. HTML extraction — content stripped from web pages; navigation, ads, and metadata removed
2. Language identification — language codes assigned per document
3. Structured headers — author/era/language metadata prepended in consistent format
4. Chunking — long works split at paragraph boundaries into 128–2,048 token segments
5. Deduplication — near-duplicate passages removed via hash-based deduplication
6. Train/val split — 94.8% / 5.2% split preserving author distribution

Quality Considerations

• Public domain: All included texts are either explicitly public domain (pre-1928 Russian publications, ancient Church Fathers) or were published on platforms with non-commercial research terms. Users should verify licensing for their specific use case.
• OCR artifacts: Some texts digitized from physical books may contain minor OCR errors, particularly in pre-1918 Russian orthography (ер, ять, і decimal).
• No annotation: This is a raw pre-training corpus; no labels, ratings, or instruction annotations are included.

---

Use Cases

Intended uses:

• Continued pre-training or domain adaptation of language models for Orthodox theology
• Multilingual (Russian/English) theological NLP research
• Language modeling and perplexity evaluation on patristic Russian
• Computational study of patristic literature, theological concepts, and Orthodox tradition
• Foundation for instruction-tuned theological Q&A systems

Not intended for:

• Training models to impersonate specific living clergy or theologians
• Generating text that falsely attributes statements to Church Fathers
• Any purpose that misrepresents Orthodox teaching or uses the corpus to oppose or parody the faith

---

Ethical Considerations

Religious Sensitivity

This corpus contains sacred and venerated texts of the Orthodox Christian tradition. Users are encouraged to engage with the material with appropriate respect.

Bias

The corpus reflects the Russian Orthodox tradition and its reception of patristic literature. Other Orthodox traditions (Greek, Coptic, Ethiopian, Serbian, Antiochian) are underrepresented. Western Christian and non-Christian theological perspectives are not included.

Authenticity

Models trained on this corpus may generate plausible-sounding theological content that does not reflect actual Church teaching. All model outputs should be verified against authoritative sources and reviewed by qualified clergy or theologians before any pastoral or catechetical application.

---

Planned Additions

A liturgical supplement is in preparation, consisting of ~11.7M additional tokens from:

Source	Content	Tokens
King James Bible	Old and New Testaments	~1.1M
Russian Synodal Bible	Old and New Testaments	~1.0M
Menologion (St. Dmitry of Rostov)	~1,200 full lives of saints	~8M
Horologion	Complete Book of Hours	~400K
Octoechos	All 8 tones hymnography	~200K
Divine Liturgy (OCA)	Full liturgy text in English	~50K
Typikon	Liturgical rubrics	~50K
Prayer Book	Orthodox prayer collections	~50K

This will be released as orthodox-patristic-corpus-v2 upon completion of the training pipeline.

---

Citation

@dataset{orthodox_patristic_corpus_2026,
  title     = {Orthodox Patristic Corpus: A Pre-Training Dataset for Orthodox Christian Theology},
  author    = {Justin Fursov},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/jayfurzy/orthodox-patristic-corpus},
  note      = {Released on the Feast of the Triumph of Orthodoxy, 2026}
}

If you train a model on this dataset, please also cite:

@misc{paterikon3b2026,
  title  = {Paterikon-3B: A Domain-Adapted Language Model for Orthodox Christian Patristics},
  author = {Justin Fursov},
  year   = {2026},
  url    = {https://huggingface.co/jayfurzy/paterikon-3b}
}

---

Dataset Card Authors

Justin Fursov

• Email: justin0106@pm.me
• LinkedIn: linkedin.com/in/justinfursov
• GitHub: github.com/jayfurz
• HuggingFace: huggingface.co/jayfurzy

Primary data source: Azbyka.ru — the Azbuka Viry ("Alphabet of Faith") Orthodox library.

---

Господи, Иисусе Христе, Сыне Божий, помилуй мя грешнаго. "Lord Jesus Christ, Son of God, have mercy on me, a sinner."