Source Library Atlas

How a map is made

A theme map is a survey of the entire catalogue for one question, organised by tradition and by time, with every held book linked and every absence named. Here is the pipeline, what the colours mean, and where it goes next.

  1. SnapshotThe full catalogue (113,276 records) is dumped to a local file: titles in every language, author, edition year, page counts, transcription and translation counts, collection tags and AI summaries.
  2. SurveySeven researchers, one per tradition, work through a checklist of canonical works, searching the catalogue in the sources' own vocabulary: original titles, transliterations, the titles of known translations and early editions.
  3. NormaliseEach held record is assigned to the canonical work it carries, with a role (original, translation, study, reception) and the work is given a conventional composition date. Works the library lacks are entered with a named public-domain source.
  4. DrawThe merged works are placed on a tradition × time grid. Readability is computed from the data, never asserted: a work is readable when a visible edition is at least 80% translated.

Reading the colours

  • Filled mark, readable. At least one edition is public and translated. The link opens the reading page with the original and the translation side by side.
  • Ring, held but not yet readable. The book is in the archive, often a bulk import from a partner library, but has no transcription or translation yet. Making it readable is a compute cost, not an acquisition.
  • Dashed ring, not held. A canonical work the library lacks. Each carries the best public-domain source the survey could name, flagged where the identifier could not be verified.

What the pipeline is not

It is not a search engine over the text. Every placement comes from catalogue metadata: titles, authors, summaries and tags. A work held inside an uncatalogued anthology can be reported as missing, and a mislabelled language is reproduced rather than repaired. The surveys flag these problems in their notes; fixing them is separate maintenance work.

Nothing on these pages was written to the library's database. The maps are read-only views built from a snapshot, and every number is reproducible from that snapshot.

Where it goes next

Any theme. The pipeline is theme-agnostic: a brief names the question, the canonical works and the genre vocabulary; the survey and the drawing are the same. Law, medicine, cosmology, love, death, dreams, the body, the city.

Passages, not just titles. With a full-text index of the transcribed and translated pages (6,523,837 and 5,085,389 pages respectively) a map can descend one level, to the passages where a tradition actually says the thing: the chapter of the Iḥyāʾ on remembrance, the paragraph of the Confessions on memory, the verse of the Māṇḍūkya on the fourth state. That is the difference between a map of books and a map of ideas. The experience atlas is a first working instance: passages retrieved by first-person probe sentences from the translated corpus, classified, gated on verbatim quotation, and plotted, with the denominator measured from a blind random sample of pages.

Cross-maps. The same traditions appear on every map. Overlaying them shows where a culture's theory of mind and its practice of meditation were written by the same hands, and where they were not.

Catalogue snapshot of 2026-09-09. Method text last revised 9 September 2026.