Created: 23-07-2026 · Last updated: 05-09-2026

title: Sense polysemy by dictionary toc: true

Sense polysemy by dictionary

Mean sense units per entry for the 11 CDSL dictionaries that carry structural sense marking. The remaining 33 dictionaries have no machine-readable sense boundary — this page does not invent a proxy (H817 dead end).

:::note Trust block. Source: data/sense_polysemy_per_dict.tsv (loader: observatory/site/src/data/sense_polysemy_per_dict.csv.py, read-only). Upstream: csl-atlas data/lexico/r2_h1.json (per-row source column). n = dictionaries · entries summed. Data date: 13-07-2026 (H817). Coverage ceiling: 11/44 — the other 33 lack structural sense-marking in digitised text; expanding n requires markup work, not a denser chart. :::

Sense units per entry (chronological)

Sorted by publication year — the editorial history of how densely senses were split.

How to read: Bar height = mean sense units per entry. Example 1: mw72 (1872) sits high: denser sense subdivision than many later titles. Example 2: Indigenous monolingual works near 1.0 mean nearly one sense unit per entry under this metric.

Year × density (family colour)

How to read: x = year, y = sense units per entry, colour = family. Example 1: Points climbing over the 19th century would be rising sense density. Example 2: Same-family pairs (mw72/mw, ap90/ap) show within-lineage drift.

Entry counts

Absolute entry volume behind the density ratio — a small dense dictionary is not the same editorial object as a large sparse one.

Family aggregate means

Unweighted mean of sense_units_per_entry across dictionaries in each family (each title counts once — not entry-weighted).

Dict × year scatter of entry volume

Fifth mark: log-scale entry count against year, size optional via radius from sense density.

Conclusion: Among the 11 sense-marked dictionaries, density is not a simple year trend: Benfey/Apte/mw72 sit high; several large Petersburg and indigenous titles sit near 1.0–1.2. Any org-wide “polysemy” claim must stay inside this n=11 set — the other 33 are not missing data, they are out of scope until sense markup exists.

Full table

Trust Block

• Source artifact: data/sense_polysemy_per_dict.tsv, served to this page via the read-only loader observatory/site/src/data/sense_polysemy_per_dict.csv.py; upstream per-row source column points into csl-atlas data/lexico/r2_h1.json; report in reports/sense_polysemy_per_dict.md.
• n: 11 dictionaries with structural sense marking (of 44 CDSL); entry counts summed per dictionary in the page's totals card.
• Data date: 13-07-2026 (H817 analysis); the CSV is regenerated from the TSV at every site build, so page and source cannot drift.
• Evidence: observed — sense-unit counts are read directly from digitised structural sense markup; no proxy is invented for the 33 dictionaries that lack machine-readable sense boundaries (H817 dead end, documented).
• Limitations: coverage ceiling 11/44 — dictionaries without structural sense markup are absent by construction, so cross-dictionary means describe only the sense-marked subset; expanding n requires markup work, not a denser chart.
• Validation: npm run build (Observable Framework) exits 0, regenerating sense_polysemy_per_dict.csv from the loader; totals card renders n and the entry sum live from the loaded CSV.
• Owner repo: csl-observatory
• Next use: revisit if any of the 33 unmarked dictionaries gains structural sense markup; otherwise stable — the census of the unmarked tail is the L1 Lexicon census's job.

Download source TSV: sense_polysemy_per_dict.tsv · report: reports/sense_polysemy_per_dict.md · sibling census: L1 Lexicon.

Dr. Mārcis Gasūns