Cross-dictionary comparison

The atlas's original core comparison: nine CDSL dictionaries on a format-robust common-block vocabulary (blocks detectable regardless of each dict's markup). It shows source-citation density and the population of each structural block, per dictionary. For size, block mass, entry lengths, and all CDSL v02 dictionaries, use the all-dictionary coverage tool. Data: the cross-dict audit in the MWS docs-pass branch.

Trust Block

• Evidence: src/data/cross-dict.json and the linked MWS cross-dictionary audit.
• Limitations: this page compares recoverable common blocks, not full semantic content or every dictionary convention.
• Validation: checked by npm run build; source data remains a compact committed comparison artifact.
• Owner repo: csl-atlas.
• Next use: inspect highlighted rows, then open exact dictionary source records before citing the pattern.

Source-citation density (<ls> per record)

PWG is ~4× denser per entry than MW in tagged <ls> citation. PWK is also a seven-part Petersburg work, but its compact editorial design drops most of PWG's named-kosha apparatus. The Sanskrit-Sanskrit lexica (SKD, VCP) and Cappeller carry no <ls> apparatus at all.

Common-block population (% of entries)

The L. hedge and info (digitisation) rows are lit for MW alone; citation is near-universal in PWG; the Sanskrit-Sanskrit lexica SKD/VCP show only head + body, marking their different genre.

Per-dictionary summary

Reading the comparison

• Block economy is a structural shape, not merely a volume count. MW, PWG, PWK, AP, WIL, Benfey, and Cappeller all reuse a small modal kernel (2–5 blocks) with a long tail. PWG is citation-denser; PWK shows that a seven-part work can still be editorially compact. See PAPER.md §9.3.
• The L. hedge is MW-specific. Of all nine, only MW carries the generic lexicographer-hedge (14% of entries); PWG cites named koshas instead. See the D2 audit.
• info is a digitisation trace. The near-universal MW info block (96%) has no analogue elsewhere — it is the infrastructure layer added by CDSL, not part of the print dictionary.
• SKD/VCP are a different genre. Both are seven-volume Sanskrit-Sanskrit works with no <lex>/<ls>; they mark gender inline and cite via inline iti <source>. The structured-bilingual block apparatus does not transfer cleanly to them, but the coverage tool records their partial fit and size.

Static figures: cross-dict-density · cross-dict-blocks.

Source: CDSL 2026-05-24 · CC-BY-SA-4.0