Datasets:

OpenChristianDataOrg
/

open-christian-data

@@ -1,187 +1,173 @@
----
-license:
-- cc0-1.0
-language:
-- en
-task_categories:
-- text-retrieval
-- question-answering
-- feature-extraction
-size_categories:
-- 100K<n<1M
-configs:
-- config_name: bible_text
-  data_files:
-  - split: train
-    path: data/bible_text.jsonl
-- config_name: catechism_qa
-  data_files:
-  - split: train
-    path: data/catechism_qa.jsonl
-- config_name: church_fathers
-  data_files:
-  - split: train
-    path: data/church_fathers.jsonl
-- config_name: commentary
-  data_files:
-  - split: train
-    path: data/commentary.jsonl
-- config_name: devotional
-  data_files:
-  - split: train
-    path: data/devotional.jsonl
-- config_name: doctrinal_document
-  data_files:
-  - split: train
-    path: data/doctrinal_document.jsonl
-- config_name: prayer
-  data_files:
-  - split: train
-    path: data/prayer.jsonl
-- config_name: reference_entry
-  data_files:
-  - split: train
-    path: data/reference_entry.jsonl
-- config_name: sermon
-  data_files:
-  - split: train
-    path: data/sermon.jsonl
-- config_name: structured_text
-  data_files:
-  - split: train
-    path: data/structured_text.jsonl
-- config_name: topical_reference
-  data_files:
-  - split: train
-    path: data/topical_reference.jsonl
----
-# Open Christian Data
-Public domain Christian literature as structured, machine-readable data — for developers and AI training.
-Commentary data is trapped in HTML and PDFs. No structured commentary dataset exists on HuggingFace. No per-chapter commentary JSON exists on GitHub.
-This dataset processes public domain Christian literature — commentaries, church fathers, confessions, catechisms, devotionals, prayers, sermons — into clean, schema-validated records with full provenance metadata.
-**GitHub:** [OpenChristianData/open-christian-data](https://github.com/OpenChristianData/open-christian-data)
-## Schema types
-| Config | Records | Description |
-|---|---|---|
-| `commentary` | 109,774 | Verse-level commentary (Matthew Henry, Barnes, Calvin, Wesley, Adam Clarke, Gill, JFB, KD, Treasury of David) |
-| `church_fathers` | 70,191 | Patristic quotes indexed by scripture reference — 325 authors (Augustine, Chrysostom, Jerome, Origen, Aquinas...) |
-| `bible_text` | 31,086 | Berean Standard Bible — all 66 books, 31,086 verses, CC0 |
-| `structured_text` | 13,207 | Paragraph-level blocks from 11 works (Calvin's Institutes, Augustine's Confessions, Pilgrim's Progress, City of God, Chesterton, Thomas à Kempis, MacDonald, Underhill, Milton, Luther's Large Catechism) |
-| `topical_reference` | 5,945 | Nave's Topical Bible — 5,322 topics, 76,957 scripture references |
-| `reference_entry` | 11,145 | Easton's, Smith's, Hitchcock's Bible Dictionaries + Torrey's New Topical Textbook |
-| `catechism_qa` | 3,279 | Question-and-answer catechisms (Westminster Shorter/Larger, Heidelberg, Baltimore #1–3, Luther's Small, Keach's, and more) |
-| `doctrinal_document` | 1,331 | Confessions and creeds at clause level (Westminster, Nicene, Chalcedonian, Belgic, Dort, London Baptist 1689, Savoy, and 30+ more) |
-| `devotional` | 1,464 | Spurgeon's Morning and Evening (732 entries) + Daily Light on the Daily Path (732 entries) |
-| `sermon` | 36 | George MacDonald's Unspoken Sermons — 3 series, 36 sermons, 171k words |
-| `prayer` | 191 | BCP 1662 Collects (85), BCP 1928 Collects (102), Didache Prayers (4) |
-| **Total** | **247,649** | |
-## Data format
-Every record is a flat JSON object. Six fields are inlined from the source file's metadata:
-| Field | Description |
-|---|---|
-| `_source_id` | Unique identifier for the source work |
-| `_source_title` | Title of the source work |
-| `_author` | Author name |
-| `_contributors` | Translators, editors, and digitizers (array, may be empty) |
-| `_schema_type` | Schema type (matches the config name) |
-| `_license` | License identifier (`cc0-1.0` or `public-domain`) |
-| `_source_url` | Canonical URL for the source |
-The remaining fields are schema-specific. All verse references use OSIS format (`Gen.1.1`, `Rom.9.1-Rom.9.5`).
-### Commentary record example
-```json
-{
-  "_source_id": "matthew-henry-complete",
-  "_source_title": "Matthew Henry's Complete Commentary",
-  "_author": "Matthew Henry",
-  "_schema_type": "commentary",
-  "_license": "public-domain",
-  "_source_url": "https://bible.helloao.org",
-  "entry_id": "matthew-henry-complete.Ezek.1.1-3",
-  "book": "Ezekiel",
-  "book_osis": "Ezek",
-  "chapter": 1,
-  "verse_range": "1-3",
-  "verse_range_osis": "Ezek.1.1-Ezek.1.3",
-  "verse_text": "In the thirtieth year...",
-  "commentary_text": "The circumstances of the vision which Ezekiel saw...",
-  "word_count": 2042
-}
-```
-### Structured text record example
-```json
-{
-  "_source_id": "pilgrims-progress",
-  "_source_title": "The Pilgrim's Progress",
-  "_author": "John Bunyan",
-  "_schema_type": "structured_text",
-  "_license": "public-domain",
-  "_source_url": "https://github.com/standardebooks/john-bunyan_the-pilgrims-progress",
-  "work_id": "pilgrims-progress",
-  "work_kind": "allegory",
-  "section_type": "part",
-  "section_label": "The First Part",
-  "section_title": "",
-  "section_path": ["The First Part"],
-  "block_index": 0,
-  "text": "As I walked through the wilderness of this world..."
-}
-```
-## Usage
-```python
-from datasets import load_dataset
-# Load a specific schema type
-commentary = load_dataset("OpenChristianDataOrg/open-christian-data", "commentary")
-church_fathers = load_dataset("OpenChristianDataOrg/open-christian-data", "church_fathers")
-catechisms = load_dataset("OpenChristianDataOrg/open-christian-data", "catechism_qa")
-# Filter by source
-matthew_henry = [r for r in commentary["train"] if r["_source_id"] == "matthew-henry-complete"]
-# All commentary on a specific verse
-rom_8_28 = [
-    r for r in commentary["train"]
-    if "Rom.8.28" in (r.get("verse_range_osis") or "")
-]
-```
-## Sources
-- **Bible text**: [Berean Standard Bible](https://berean.bible) — CC0 since April 2023
-- **Commentary**: [HelloAO Bible API](https://bible.helloao.org) — Matthew Henry, JFB, Gill, Adam Clarke, KD; [CrossWire SWORD](https://www.crosswire.org/sword/) — Barnes, Calvin, Wesley, Treasury of David; all public domain
-- **Church Fathers**: [HistoricalChristianFaith/Commentaries-Database](https://github.com/HistoricalChristianFaith/Commentaries-Database) — 325 authors, public domain
-- **Devotionals**: Christian Classics Ethereal Library (ccel.org) — Spurgeon's Morning and Evening (ThML XML); CrossWire SWORD — Daily Light on the Daily Path; public domain
-- **Structured texts**: [Standard Ebooks](https://standardebooks.org) — 9 titles in CC0-annotated XHTML; Project Gutenberg — Luther's Large Catechism, Calvin's Institutes, Augustine's Confessions; public domain
-- **Catechisms**: Project Gutenberg — Luther's Small Catechism, Baltimore Catechisms #1–3; additional catechisms (Westminster, Heidelberg, Keach's, and more); public domain
-- **Confessions & creeds**: Westminster Confession, Nicene Creed, Chalcedonian Definition, and 30+ more historic documents; public domain
-- **Prayers**: [eskimo.com BCP 1662](https://eskimo.com/~lhowell/bcp1662/) (Lynda M. Howell); [episcopalnet.org BCP 1928](https://www.episcopalnet.org/1928bcp/); [Wikisource Didache](https://en.wikisource.org/wiki/Didache_(Lake_translation)) (Kirsopp Lake 1912 translation); public domain
-- **Bible dictionaries**: Easton's (1893), Smith's (1863), Hitchcock's (1874), Torrey's (1897); public domain
-- **Topical reference**: [CrossWire SWORD](https://www.crosswire.org/sword/) — Nave's Topical Bible (1896); public domain
-- All authors died before 1928; texts are unambiguously public domain.
-## Attribution
-Devotional text sourced from the Christian Classics Ethereal Library (ccel.org). ThML parsing permitted per correspondence with CCEL (April 2026).
-## License
-This dataset is released under [CC0 1.0 Universal](https://creativecommons.org/publicdomain/zero/1.0/) — dedicated to the public domain. You can use it for any purpose, including commercial AI training, without restriction or attribution.
-The underlying texts are public domain works by authors who died before 1928. The pipeline code, schemas, and tooling used to produce this dataset are licensed separately under CC BY-NC 4.0.

+---
+license: cc0-1.0
+language:
+- en
+pretty_name: Open Christian Data
+size_categories:
+- 100K<n<1M
+task_categories:
+- text-generation
+- question-answering
+- text-classification
+- feature-extraction
+tags:
+- christian
+- theology
+- bible
+- commentary
+- sermon
+- catechism
+- confession
+- church-fathers
+- public-domain
+- religion
+configs:
+- config_name: bible_text
+  data_files:
+    - path: data/bible_text.jsonl
+      split: train
+- config_name: catechism_qa
+  data_files:
+    - path: data/catechism_qa.jsonl
+      split: train
+- config_name: church_fathers
+  data_files:
+    - path: data/church_fathers.jsonl
+      split: train
+- config_name: commentary
+  data_files:
+    - path: data/commentary.jsonl
+      split: train
+- config_name: devotional
+  data_files:
+    - path: data/devotional.jsonl
+      split: train
+- config_name: doctrinal_document
+  data_files:
+    - path: data/doctrinal_document.jsonl
+      split: train
+- config_name: hymn_collection
+  data_files:
+    - path: data/hymn_collection.jsonl
+      split: train
+- config_name: prayer
+  data_files:
+    - path: data/prayer.jsonl
+      split: train
+- config_name: reference_entry
+  data_files:
+    - path: data/reference_entry.jsonl
+      split: train
+- config_name: sermon
+  data_files:
+    - path: data/sermon.jsonl
+      split: train
+- config_name: structured_text
+  data_files:
+    - path: data/structured_text.jsonl
+      split: train
+- config_name: topical_reference
+  data_files:
+    - path: data/topical_reference.jsonl
+      split: train
+---
+# Open Christian Data
+A free, author/work-led dataset of public domain Christian literature — Bible text, commentaries, sermons, catechisms, confessions, theology, prayer, and reference works, projected into clean records with full provenance metadata. Intentionally designed for developers and AI/LLM training.
+All content is written by authors who died before 1928. All texts are in the public domain. The dataset is CC0.
+## Export shape
+The live whole-corpus HuggingFace export is the per-schema JSONL export generated by
+`build/scripts/export_huggingface.py`. The `original`/`modernised` export path in the source
+repository is a separate NSH-style publish-projection path and is not the current whole-corpus
+HuggingFace release.
+## What's included
+The public inventory is counted by recognized work units: the title-level item a reader, library, citation, or source catalog would treat as present. Export records, verses, dictionary entries, topics, Q&A pairs, chapters, and quotations are subordinate technical units.
+Generated from current `data/` with `py -3 build/tools/count_dataset_records.py`.
+| Category | Work units | Top-level records |
+|---|---:|---:|
+| Bible text | 9 | 282,395 |
+| Catechisms | 15 | 3,509 |
+| Church Fathers quotation source | 1 | 70,164 |
+| Commentaries | 34 | 113,903 |
+| Devotionals | 2 | 1,464 |
+| Doctrinal documents | 38 | 38 |
+| Hymns | 1 | 34,904 |
+| Lexicon | 1 | 0 |
+| Prayers | 4 | 205 |
+| Reference works | 7 | 26,305 |
+| Sermons | 7 | 5,297 |
+| Structured texts | 288 | 289 |
+| Topical reference works | 1 | 5,322 |
+| **Total** | **408** | **543,795** |
+The live HuggingFace export flattens structured texts and doctrinal documents, so this release exports **803,985 JSONL rows**. The source repository includes the generated work metadata audit at `docs/WORK_CATALOG.md` and a browser-friendly review surface at `docs/WORK_CATALOG.html`.
+## Loading the data
+```python
+from datasets import load_dataset
+# Load a single config
+original = load_dataset("OpenChristianData/open-christian-data", "original")
+modernised = load_dataset("OpenChristianData/open-christian-data", "modernised")
+# Filter to a specific author
+wesley = [r for r in modernised["train"] if r["meta"]["author_slug"] == "wesley"]
+# Filter to works with no modernised sibling by reading the coverage table above
+```
+## Data structure
+Every record is a flattened row combining the resource-level metadata with entry-level fields. Key fields present on all records:
+| Field | Description |
+|---|---|
+| `_source_id` | Unique identifier for the parent resource (e.g., `matthew-henry-complete`) |
+| `entry_id` | Unique identifier for this entry (e.g., `matthew-henry-complete.Ezek.1.1-3`) |
+| `author` | Author name |
+| `tradition` | Array of theological tradition labels (e.g., `["reformed", "puritan"]`) |
+| `era` | Historical era (`patristic`, `medieval`, `reformation`, `post-reformation`, `modern`) |
+| `license` | `cc0-1.0` or `public-domain` |
+| `schema_type` | Content type — one of the config names above |
+Schema-specific fields vary by content type. Full field documentation is in the [source repository](https://github.com/OpenChristianData/open-christian-data/blob/main/docs/SCHEMA_SPEC.md).
+## Provenance
+Every record carries full provenance — the source URL, source format, download date, processing script, and a SHA-256 hash of the source file. This is stored in the `provenance_*` fields in the flattened export.
+## Verse references
+All verse references use [OSIS format](https://crosswire.org/osis/) (`Gen.1.1`, `Rom.9.1-Rom.9.5`). Where sources use non-standard reference formats, both the raw form and the normalized OSIS form are stored.
+## License
+- **Data**: [CC0 1.0 Universal](https://creativecommons.org/publicdomain/zero/1.0/) — all underlying texts are public domain; the structuring and provenance tracking are dedicated to the public domain
+- **Bible text (BSB)**: [CC0 1.0](https://berean.bible/licensing.htm)
+- **Code and schemas**: [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) — free for non-commercial use; commercial use requires a license (see [source repository](https://github.com/OpenChristianData/open-christian-data))
+## Citation
+```bibtex
+@dataset{open_christian_data_2026,
+  title        = {Open Christian Data},
+  author       = {OpenChristianData},
+  year         = {2026},
+  publisher    = {HuggingFace},
+  url          = {https://huggingface.co/datasets/OpenChristianData/open-christian-data},
+  license      = {CC0-1.0}
+}
+```
+## Source repository
+[github.com/OpenChristianData/open-christian-data](https://github.com/OpenChristianData/open-christian-data)