Thompson Motif Index Explorer

What is this?

The Thompson Motif-Index of Folk-Literature (Stith Thompson, 1955-1958) is the international standard for classifying folklore motifs across cultures — 46,230 entries spanning myths, magic, animals, death, marvels, and more. This explorer is the first systematic computational mapping between TMI and the Estonian and Finnish runosong corpora.

Data sources

TMI data: 46,230 motifs from the Trilogy TMI+ATU dataset (Hagedorn 2022, CC BY-SA 4.0), based on Thompson's original six-volume index (Indiana University Press, 1955-1958). Organized into 23 chapters (A: Myths through Z: Miscellaneous) with up to 6 levels of hierarchical subdivision.
Song types: 25,724 unique type names extracted from 160,088 typed poems across SKVR (98.6% typed), ERAB (65.5%), and KR (6.4%). Only types with ≥10 poems (4,227 types) are used for matching.
Translations: 311,488 lemma-to-English translations from lemma_english.json and 471,241 English vocabulary entries from translation_index.json.
Word embeddings: GloVe-300 (Pennington, Socher & Manning, 2014) — 400,000 English word vectors trained on Wikipedia + Gigaword. Used for semantic similarity computation at build time.

Three matching tiers

Tier 1 — Song Type Bridge (semantic): The primary matching method. Finnic song type names are translated word-by-word to English using lemma_english.json (with Finnish suffix-stripping fallbacks), then represented as GloVe-300 semantic centroids. TMI motif names are similarly vectorized. FAISS cosine similarity finds the top matching motifs per song type. This captures synonym relationships invisible to exact matching: “birth” ≈ “origin” (~0.65), “serpent” ≈ “snake” (~0.85), “enchantment” ≈ “magic” (~0.60). An exact keyword overlap bonus (+0.15) rewards direct matches like “iron” in “Origin Of Iron”.
Tier 2 — Vocabulary Overlap: Matches specific TMI keywords (those appearing in ≤100 TMI motifs) against the RunoVerse English translation index to find runosong lemmas that share vocabulary with the motif. Requires ≥2 keyword matches to avoid false positives.
Tier 3 — Culture-Area Flagging: Parses TMI notes for word-boundary mentions of “Finnish” (602 motifs), “Estonian” (231), “Livonian” (127), “Norse” (93), “Lappish” (87), “Finno-Ugric” (73), and other Nordic/Baltic terms. These are TMI motifs where Thompson himself documented connections to Finnic traditions.

How to use

By Chapter (default tab): Browse the 23 TMI chapters as an accordion hierarchy. Click a chapter to see its sections, click a section to see individual motifs. Matched motifs show gold score badges (≥0.60 = high confidence) and green FIN flags (culture-area documented). Unmatched motifs appear in gray. Click any motif to expand its detail panel showing matched song types and vocabulary.
By Song Type: Lists all 3,897 matched Finnic song types sorted by poem count. Each shows its language (FI/ET), poem count, and number of TMI motif matches. Click to expand and see individual motif matches with semantic scores. Click a motif ID to navigate to its chapter location.
Search: Type at least 2 characters to search across all 46K motif names and song type names simultaneously. Results show both song types and motifs with match indicators.

Filters

All: Show all 23 chapters.
Song Type Matches: Only chapters containing Tier 1 semantic matches.
Finnic-Documented: Only chapters with culture-area flagged motifs.
High Relevance: Only chapters most relevant to runosong tradition (A: Myths, D: Magic, E: Death, F: Marvels).

Scores & confidence

Semantic scores (0.50–1.00) indicate thematic similarity strength between a song type's English translation vector and a TMI motif's keyword vector. Gold (≥0.60): high confidence thematic match. Gray (0.50–0.60): possible match, lower confidence. Scores below 0.50 are discarded. All matches are algorithmic suggestions — researchers should validate for their specific use case.

Navigating to poems

Song type pills in motif details show the Finnic type name and poem count. To see actual poems of that type, use the Concordance page to search for the type name, or browse the Reader to find poems classified under that type. Vocabulary match pills link directly to the Lexicon entry for each lemma.

URL deep linking

Share specific views: ?chapter=A (open chapter), ?motif=A1414 (navigate to motif), ?type=Raudan+synty (song type lookup), ?q=wolf (search), ?tab=types (tab selection).

Limitations

GloVe centroids can produce false positive matches between unrelated concepts. 1.3% of song types have no translatable keywords. Finnish compound words may not decompose fully. The “tuli” lemma maps to both “fire” and “come” due to homonymy. See project documentation for full methodology details.

Attribution

TMI data: Trilogy TMI+ATU (Hagedorn 2022), CC BY-SA 4.0. Original: Thompson, S. (1955-1958). Motif-Index of Folk-Literature. Embeddings: GloVe (Pennington et al. 2014). See also: MOMFER (Karsdorp et al. 2015, Meertens Institute).