RUNA-1: a typed biosemiotic knowledge-graph embedding for European ecology (v1.3.0)

RUNA-1 is a typed knowledge-graph embedding (PyKEEN, model: BoxE, embedding_dim 128) over open European ecological data, in which species, discretized environmental-state nodes, detected-community nodes and signal/element nodes share one geometric space where proximity encodes ecological and biosemiotic relatedness.
On top of conventional ecological relations (predation, pollination, mycorrhizae, parasitism, etc., mapped where possible to the OBO Relations Ontology), it adds biosemiotic sign-relations trained as ordinary typed edges: indicatorOf (a species as a sign of an environmental state, from EIVE/Ellenberg) and keystoneSignProducerIn (keystone sign-producer within a detected community).
New in v1.3.0 — the signal-web layer (the perceivesSignal slot, finally built). Following the relational semiosis of Kohn's How Forests Think, v1.3 adds two relations: castsSignal (which species produce which kinds of signal — sound types song/call/alarm/flight-call/echolocation, from xeno-canto recordings) and readsSignal (which species read the elements as signs — the magnetic field and electric fields, hand-curated from the confirmed-taxa literature, e.g. European robin/eel/salmon reading the magnetic field; lampreys/sturgeon/wels catfish/sharks reading electric fields). Every signal edge is provenance-typed (MEASURED at species level vs confirmed-at-group level).
v1.2 carried forward: a noise-cleaned base plus 96 place-based dependency edges curated by the nine agnt eco agents and fact-checked by a multi-model consensus (claude-opus-4-8 and gpt-5.4 reading the cited sources; included only where Claude confirmed and GPT did not refute).
Priority claim: the first trained relational embedding that operationalizes biosemiotic relations for ecology.
Validation (honest scope, v1.3.0): held-out link prediction filtered MRR 0.301 (Hits@10 0.50). On the independently-validated indicator axes (model placement vs real GBIF occurrence × environment, non-circular): temperature Spearman rho = 0.602 vs CHELSA bio1 and soil pH rho = 0.340 vs SoilGrids (above the expert-input ceiling of 0.213). A same-data benchmark shows BoxE beats a frequency baseline ~10×, ComplEx clearly, and ties RotatE on link prediction — but RotatE places the thermal axis at only rho 0.114 because it cannot encode the 1-to-N indicator relations BoxE handles. Moisture and nutrients remain unvalidated (proxies invalid); keystoneSignProducerIn and the new signal layer are learnable but not yet independently validated.
Contents: the frozen reconciled + agent-augmented + signal-layer triple set, the relation schema, the full derivation/verification code, the trained BoxE model (v1.3), the multi-axis validation data, a full benchmark (BENCHMARK.md), documentation, and a SHA-256 manifest. Derived from GLOBI, Mangal, EIVE 1.0, GBIF, CHELSA, SoilGrids, xeno-canto, plus place-based agent research with cited primary sources.