Skip to main content

Created: 07-07-2026 · Last updated: 06-11-2026

What Each Dictionary Quotes

A dictionary's authority comes from its citations: the classical texts it quotes to attest each sense. The Cologne sources mark these with <ls> tags (see abbreviations & citations); the sibling csl-atlas project resolved each dictionary's own abbreviations (MBh., MBH, Mahābhārata) to shared canonical text nodes, making citation habits comparable across dictionaries.

Pick a dictionary to see its most-quoted texts, or keep All dictionaries for the corpus-wide view.

Top 15 of 874 canonical texts, by total resolved <ls> citations across the 11 covered dictionaries (826,752 resolved citations).

Mahābhārata56,822 · in 8 dicts
Ṛgveda39,025 · in 11 dicts
Rāmāyaṇa38,154 · in 9 dicts
Manusmṛti28,140 · in 8 dicts
Bhāgavata-Purāṇa24,300 · in 8 dicts
Raghuvaṃśa24,025 · in 8 dicts
Aṣṭādhyāyī (Pāṇini)21,791 · in 3 dicts
Śabdakalpadruma20,232 · in 7 dicts
Abhidhānacintāmaṇi18,073 · in 3 dicts
Indische Sprüche18,030 · in 4 dicts
Kathāsaritsāgara17,015 · in 9 dicts
Amarakoṣa14,918 · in 8 dicts
Pañcatantra14,853 · in 7 dicts
Harivaṃśa13,685 · in 7 dicts
Medinīkośa13,246 · in 6 dicts
Data table
TextCitationsDictionaries citing it
Mahābhārata56,8228
Ṛgveda39,02511
Rāmāyaṇa38,1549
Manusmṛti28,1408
Bhāgavata-Purāṇa24,3008
Raghuvaṃśa24,0258
Aṣṭādhyāyī (Pāṇini)21,7913
Śabdakalpadruma20,2327
Abhidhānacintāmaṇi18,0733
Indische Sprüche18,0304
Kathāsaritsāgara17,0159
Amarakoṣa14,9188
Pañcatantra14,8537
Harivaṃśa13,6857
Medinīkośa13,2466

How to read it. PWG dominates the graph (536,172 resolved citations — Böhtlingk & Roth cited on an industrial scale); its profile is epic-heavy (Mahābhārata, Rāmāyaṇa) with Pāṇini's Aṣṭādhyāyī third — a grammarians' dictionary. AP's top text is the Raghuvaṃśa: Apte built for readers of kāvya. BHS's citations are almost entirely Buddhist texts no other dictionary touches. The Vedic concordances (VEI, PUI) don't appear here at all — they carry zero <ls> tags (their citation systems are structural hymn/verse references), and Grassmann's Rigveda references are verse numbers rather than resolvable text abbreviations.

Trust block
  • Evidence: src/data/citation-sources.json, vendored by scripts/build-atlas-viz.mjs from the csl-atlas citation graph (ls_citation_edges.tsv + ls_citation_nodes.tsv, method in its README): 826,752 resolved <ls> citations → 874 canonical texts across 11 dictionaries (1699 edges).
  • Limitations: coverage is the 11 dictionaries whose abbreviation keys resolve — not all 45. Resolution quality varies: MW's 312k <ls> tags currently resolve to only 5 coarse nodes (its grammatical markers are filtered, most literary abbreviations are still unresolved), and MD to 4 — read those two as placeholders, not profiles. Canonicalization is imperfect across dictionaries (MD's "Rigveda" is not yet folded into "Ṛgveda"). Counts are citation frequency, not importance.
  • Validation: vendored totals match the csl-atlas edge list exactly (826,752); both TSVs' SHA-256 are pinned in the feed, and npm run check:atlas-viz re-derives the projection from the sibling checkout and fails on any drift. The per-dictionary top texts were spot-checked against the TSV.
  • Owner repo: csl-atlas (data) / this repo (rendering).
  • Next use: if a dictionary you use is missing here, its abbreviation key is the gap — see the csl-atlas unresolved-keys worklist.

See also​

  • The citation graph explained — what this graph is, how it was built, how to query it
  • Abbreviations & citations — how <ls> tags work
  • The dictionary landscape — era × novelty × size for all 43
  • Corpus attestation — how often each word is actually used
  • Abbreviations comparison — the per-dictionary legends the resolver uses