ciscoriordan/dilemma
Rebuild with the expanded Patrologia Graeca evidence
Upload data/form_profile.db
Rebuild after the Diorisis elision and paradigm generator fixes
Rebuild after the Diorisis elision and paradigm generator fixes
Rebuild after the Diorisis elision and paradigm generator fixes
Upload data/form_profile.db
Remove editorial sigla from GLAUx pairs
Upload data/spell_index.db
Upload data/lookup.db
Upload data/spell_index.db
Upload data/lookup.db
Upload data/spell_index.db
Upload data/lookup.db
Upload data/spell_index.db
Upload data/lookup.db
Upload data/spell_index.db
Upload data/lookup.db
Upload data/spell_index.db
Upload data/lookup.db
Upload data/spell_index.db
Upload data/lookup.db
Upload data/pedalion_pairs.json with huggingface_hub
Upload data/tagnt_pairs.json with huggingface_hub
Upload data/ptnk_pairs.json with huggingface_hub
Upload data/treebank_pos_lookup.json with huggingface_hub
Upload data/lemma_attestation.json with huggingface_hub
Upload data/spell_index.db with huggingface_hub
Upload data/lookup.db with huggingface_hub
Upload tagger/grc/tagger_model.pt with huggingface_hub
Upload tagger/grc/tagger_labels.json with huggingface_hub
Upload tagger/grc/tagger.onnx with huggingface_hub
grc-s3: retrained on PROIEL/Gorman-purged pairs (grc-mode benchmarks improved)
spell_index: rebuilt
lookup.db: Ionic indefinite family + hews/hepta/oikia homograph fixes
combined-s3: retrained on PROIEL/Gorman-purged glaux_pairs (benchmarks unchanged)
form_citations: +OGA CTS loci
form_profile: +OGA evidence (795 newly covered works)
lemma_attestation: OGA third source via cog oga-v1 (303K lemmas / 34M tokens)
glaux_pairs: PROIEL/Gorman-manual purged
form_citations: PROIEL/Gorman-manual purged
form_profile: PROIEL/Gorman-manual purged
corpus_freq: rebuilt
lemma_attestation: PROIEL/Gorman-manual purged
spell_index: rebuilt
lookup.db: PROIEL/Gorman-derived GLAUx trees purged
spell_index.db: rebuilt from Gorman-free lookup
lookup.db: Gorman held out as gold corpus + bare-elided-stem overrides
spell_index.db: rebuilt from canonicalized lookup
lookup.db: elided-key canonicalization + corrected entries (sigma/pote pronouns, U+1FBD keys)
Remove stale *_commercial twins (builds are openly-licensed by default; these doubled the base download)