catena / README.md
TheAlvaroBalbin's picture
Add the Roman Catechism (Catechism of the Council of Trent), 929 sections
8978579 verified
|
Raw
History Blame Contribute Delete
5.65 kB
---
license: other
license_name: public-domain
language:
- en
- la
tags:
- theology
- catholic
- christianity
- aquinas
- summa-theologica
- church-fathers
- catena-aurea
- patristics
- bible
- vulgate
- citations
- knowledge-graph
- retrieval
pretty_name: "Catena: an open, cited corpus of the Catholic tradition"
size_categories:
- 10K<n<100K
configs:
- config_name: summa
data_files: summa.jsonl
- config_name: catena_aurea
data_files: catena_aurea.jsonl
- config_name: roman_catechism
data_files: roman_catechism.jsonl
- config_name: douay_rheims
data_files: douay_rheims.jsonl
- config_name: clementine_vulgate
data_files: clementine_vulgate.jsonl
- config_name: scripture_edges
data_files: scripture_edges.jsonl
- config_name: father_edges
data_files: father_edges.jsonl
- config_name: internal_edges
data_files: internal_edges.jsonl
---
# Catena: an open, cited corpus of the Catholic tradition
A clean, machine-readable, **public-domain** corpus of the Catholic tradition where
every unit of text carries a **canonical citation**, the text is stored **verbatim**
(never paraphrased or truncated), and the tradition's own **citation graph** is
included as loadable data. Built for grounded, cite-or-refuse retrieval.
Source, ingest code, a live web explorer, and an MCP grounding server:
**https://github.com/AlvaroBalbin/catena**
## Subsets
| config | rows | what |
|--------|------|------|
| `summa` | 3,115 | The Summa Theologica of Thomas Aquinas, one row per article, with its canonical citation (`ST I, q.2, a.3`) and full verbatim text. |
| `catena_aurea` | 814 | The Catena Aurea - Aquinas's patristic "golden chain" on the four Gospels, one row per verse-pericope: the verse commented on, the Church Fathers in the chain, and their full verbatim commentary. |
| `roman_catechism` | 929 | The Roman Catechism (Catechism of the Council of Trent), one row per subsection, with its Part and heading and full verbatim text. Public-domain McHugh &amp; Callan translation (1923). |
| `douay_rheims` | 35,786 | The Douay-Rheims Bible (Challoner revision), one row per verse, English. |
| `clementine_vulgate` | 35,809 | The Clementine Vulgate (Sixto-Clementine, 1592), one row per verse, Latin - keyed identically to the Douay-Rheims so a citation resolves to both languages. |
| `scripture_edges` | 9,449 | The citation graph: one row per (Summa article -> Scripture verse) edge. |
| `father_edges` | 3,783 | The patristic golden chain as edges: one row per (Catena Aurea pericope -> Gospel verse it comments on), with the Fathers who speak - which Fathers weigh in on which verse. |
| `internal_edges` | 3,775 | The citation graph: one row per (Summa article -> Summa article) edge. |
```python
from datasets import load_dataset
summa = load_dataset("TheAlvaroBalbin/catena", "summa")
fathers = load_dataset("TheAlvaroBalbin/catena", "catena_aurea")
verses = load_dataset("TheAlvaroBalbin/catena", "douay_rheims")
latin = load_dataset("TheAlvaroBalbin/catena", "clementine_vulgate")
edges = load_dataset("TheAlvaroBalbin/catena", "scripture_edges")
```
Every verse shares a `verse_key` (`john/1/14`) across the two Bibles and the edge
tables, so the subsets join cleanly: an article's Scripture citations resolve to real
verse text in English and Latin.
## The three disciplines
1. **Lossless.** Text is captured verbatim from the source edition; only whitespace is
normalized. Nothing is summarized or silently truncated.
2. **Addressable.** Every unit has a stable, canonical citation and a join key.
3. **Grounded.** The reference demo (in the GitHub repo) answers only from retrieved
units, always cites, and refuses when the corpus does not contain the answer.
## Sources and rights
All texts are public domain by age and are redistributed with their edition attributed.
Full provenance and the per-text rights reasoning are in
[`data/SOURCES.md`](https://github.com/AlvaroBalbin/catena/blob/main/data/SOURCES.md).
- **Summa Theologica** - trans. Fathers of the English Dominican Province, 2nd rev. ed.
(1920-22). Public domain.
- **Catena Aurea** - Aquinas's compilation of the Church Fathers on the Gospels, in the
Oxford translation edited by J. H. Newman (1841-45). Public domain.
- **Roman Catechism** (Catechism of the Council of Trent) - tr. J. A. McHugh & C. J.
Callan (1923); public domain in the US (pre-1929). A different work from the
copyrighted modern Catechism, which is not included.
- **Douay-Rheims Bible** - Challoner revision (1749-52), via Project Gutenberg #1581.
Public domain.
- **Clementine Vulgate** - Sixto-Clementine (1592); electronic edition by The Clementine
Vulgate Project (ed. M. Tweedale, 2005), released to the public domain.
Copyrighted texts (the modern Catechism, modern Bible or encyclical translations) are
**not** included. The `license: other` tag denotes public-domain-by-age; there is no
rights holder to waive rights, and none is implied.
## Citation graph coverage
96.8% of the Summa's Scripture citations parse (8,491 of 8,768). Because a single
citation to a verse range or a whole chapter touches several verses, those expand to
9,449 verse-level `article -> verse` edges here (`scripture_edges`); chapter-level
citations and the full adjacency maps are in `graph/`. 99% of the Summa's internal
cross-references resolve into 3,775 `article -> article` edges (`internal_edges`).
The graph builder and coverage details are in the GitHub repository.