DeepSeek Commentary Search … Per-Lemma Verse Commentary Both ? Help

×

About this page

Two data sources

This page provides keyword search over two distinct DeepSeek commentary datasets, selected via the toggle at the top:

1. Per-Lemma Comments (default)

Short notes generated per wordform. Indexed at build time into a monolithic JSON file.

• 24,149 unique keywords (all lowercased)
• 1,063,284 indexed keyword → wordform mappings (ET 617,333 · FI 445,951)
• 2,445,999 total comments
• Index build date: …

2. Verse Commentary

Full-verse philological commentary — longer, richer analysis per verse line. Indexed into a sharded search index (keyword catalog + 64 postings shards loaded on demand).

• ~425K keywords (after MIN_OCCURRENCES=2 filter)
• 4.3M verse commentaries across 289,866 poems
• Postings are sampled (cap 10/keyword, 120-char truncation) — use the Poem Dossier tool for full uncapped commentary per poem
• The ASCII-only tokenizer indexes English keywords from the commentary text; quoted Finnic word forms are not indexed as keywords

Tier guide (per-lemma mode)

• ≥500 — generic linguistic tags: dialectal, form, archaic
• 21–500 — specific morphology, common semantic fields
• 6–20 — narrow topics, mid-tail vocabulary
• 2–5 — rare tail, often the most interesting: zebra, youthful

URL deep links

• ?q=partitive — keyword detail view (per-lemma)
• ?q=partitive&w=kannel — highlight wordform row
• ?source=verse&q=dialectal — verse commentary keyword
• ?wf=kannel — Tool 1 dossier
• ?a=partitive&b=dialectal — Tool 2 co-occurrence