SOURCES.md

# Sources — islamlab, The Islamic Sciences Corpus

islamlab is a volunteer effort. Our aim is to present these sources carefully
and at high quality, and to help them travel faithfully — in the spirit in
which they were written — not to claim them as ours.

None of this is ours. Every work here belongs to the scholar who wrote it and
to the people who took the trouble to make it readable by a machine — often
quietly, for years, without being asked. What we did was select, sort, segment
and check. What follows is who to thank.

## The projects this corpus rests on

- **[Open Islamicate Texts Initiative (OpenITI)](https://openiti.org)** — The
  machine-readable corpus of premodern Islamicate texts, pinned here to release
  **2025.1.9**, licensed CC BY-NC-SA 4.0. OpenITI is led from the Aga Khan
  University (ISMC), the University of Vienna, Leipzig University and the
  Roshan Institute at the University of Maryland, with support from the KITAB
  project and the Qatar National Library.
- **[al-Maktaba al-Shāmila](https://shamela.ws)**, reached through the
  [AuthenticIlm](https://huggingface.co/datasets/AuthenticIlm/Shamela4_Full_DB)
  packaging of its library (MIT). Shamela is where a great many of these works
  were first typed and proof-read, over decades, by people who did not sign
  their work.
- **[Rasaif](https://rasaif.com/)** — The sentence-aligned Arabic–English
  translations behind the `parallel_en` config (CC BY-4.0).

Which of the two libraries a given text reached us through is recorded per row
in the `source` column of `works`, and nowhere else: it is a fact about our
copy, not about the book.

The works themselves are older than any of us and belong to their authors —
al-Ghazālī, al-Māwardī, Ibn ʿAsākir, Ibn Manẓūr, al-Sarakhsī, al-Dhahabī and
the rest. Where a printed edition stands behind a text, it is named per work in
the `edition` column of the `works` config.

## Consulted, but never quoted

Some things we read carefully and published nothing from. They told us what a
work is and which shelf it belongs on, and then we set them down. We name them
because using something and reproducing it are different acts, and both should
be said out loud.

- **Library catalogue headings** carried in the OpenITI metadata (al-Maktaba
  al-Shāmila, al-Jāmiʿ al-Kabīr, Brockelmann's *GAL*) — these decided which of
  the eight sciences a work was filed under when its genre tags were silent.
  Not one of those headings is in the data; they never became a column, a row
  or a source line.
- **[TDV İslâm Ansiklopedisi](https://islamansiklopedisi.org.tr)** — Our first
  stop when we needed to be sure who an author was, when he died, and where he
  stood. Read, never reproduced. It is what let us separate Qāḍī Abū Bakr Ibn
  al-ʿArabī (d. 543, Mālikī) from Ibn ʿArabī (d. 638), and keep the first.

## What we left out, on purpose

The gate is described in the card; the machine-readable form of it lives in
`data/banned.json` in our build repository, with a written reason beside every
exclusion. In summary: material outside Ahl al-Sunnah wa'l-Jamāʿah, the
Hellenistic philosophers, non-Muslim and orientalist authors, anything whose
author died after 1350 AH, and anything that belongs to
[`quranlab/*`](https://huggingface.co/quranlab) or
[`risaleinur/*`](https://huggingface.co/risaleinur) instead.

## If we got something wrong

If you hold rights to a work here, or we have credited someone poorly, or we
have put a scholar on the wrong side of the gate, write to us on the
**Community** tab of
[`islamlab/islamic-sciences`](https://huggingface.co/datasets/islamlab/islamic-sciences).
A removal request is honoured promptly and without argument; a correction is
welcomed with thanks.