Samudra Manthanam: Markup-Aligned Sanskrit–Russian Parallel Corpus of 148 Sources (corpus of record)
Year: 2026
Type: dataset
Abstract: Markup-aligned parallel corpus of 148 sources (574,939 segment records), extracted deterministically from the source markup — in the source HTML the Sanskrit text and its Russian translation are interleaved as sibling blocks, and the converter reads the pairing directly; no statistical alignment is run. Composition: 119 verse sources (208,230 records), 15 dictionaries (321,672), 14 prose works (45,037). Headline: 78,219 clean 1:1 verse pairs (78,139 live re-count), plus 10,009 Russian-only and 80 Sanskrit-only monolingual segments. Includes 11 Russian translations of the Bhagavadgītā (1788–2015) with three commentaries — a built-in diachronic register set. Rights (per-layer): Sanskrit source text public domain by age; Russian translations pre-1930 editions public domain, 20th–21st-century editions in copyright with a grey residual, shipped under the project's standing ruling of 2026-08-08 with translators documented per source; imported DCS lemma/morphology layer CC BY 4.0 (Hellwig), redistributed with attribution; upstream code/schema/documentation Apache-2.0. The corpus texts carry their own rights and must be cited by their print editions; full table in the data statement file. Provenance: platform and converter at https://github.com/gasyoun/SamudraManthanam (software concept DOI 10.5281/zenodo.21317315 — cite that DOI for the software, this DOI for the corpus). Post-report upstream additions (121 files / 199,379 records) are excluded from this corpus of record and may appear in a future separately-versioned release. Corpus of record is byte-stable across four re-counts (2026-06-26, 2026-07-08, 2026-07-11, 2026-08-10); every figure recomputes in one deterministic pass.
Author: Gasūns, Mārcis
Language: Sanskrit
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204586925