L3 · Corpus & usage

Statistics over running-text usage: the DCS full corpus, SamudraManthanam's Sa↔Ru parallel corpus, frequency layers, and the still-open meter/accent statistics. Part of the statistics census overview (H817 WS1.3).

:::note Trust block. Source: stats_census_register.csv, rows where layer = L3, aggregated from VisualDCS, SamudraManthanam, and kosha. n = . As of 06–13-07-2026. :::

Headline magnitudes

How to read: log-scale bar of corpus-scale counts. Example: the 40.6M stop-word parallels dwarf the 5.7M token corpus itself by an order of magnitude — it is a pairwise/derived count, not raw text volume; read it as a derived-index size, not "more text than DCS holds."

Status breakdown

Full table

Download: stats_census_register.csv (full register, all layers) · Data downloads.

Open gaps

POS distribution per text moved from ○ to closed under H817 WS1.2 (now census-registered). Meter/prosody (SanskritKaraoke) and Vedic accent coverage (pending VedaWeb reuse) remain genuinely ○ not started — scheduled Q3 (WS3.5) per the roadmap, not blocked on anything this pass could resolve.

Chart density note: 2 Plot.plot calls (magnitude bar + status bar) — justified per the same heterogeneous-units reasoning as the L1 page.