Ancient Greek, Modern Greek, and English aligned.
Tap a word in any language to see its Ancient Greek context from Cunliffe/LSJ.
Morphological analysis uses Dilemma, a diachronic Homeric-through-Modern Greek lemmatizer built for this project.
A single model spanning Homeric, Koine, Byzantine, Katharevousa, Cypriot, Cretan, and Standard Modern Greek, with orthographic normalization for papyrological and Byzantine texts. Scores 92.7% on the DBBE Byzantine gold standard.
Prose translations are redistributed to match Homer's verse using a novel adaptation of Knuth-Plass. Given N source verse lines and M translation words, dynamic programming finds cut points for
where a(w, l) = +2 if word w aligns to line l, −1 if elsewhere, 0 if unaligned.
The Riordan revision converts polytonic to monotonic, replaces Katharevousa prepositions (εἰς → σε, διά → για), standardizes Corfu dialect forms (σιμά → κοντά, ξεύρω → ξέρω), and restores patronymics that Polylas had simplified, bringing the text closer to both AG and SMG simultaneously.
Murray's 1924 prose is accurate but steeped in Wardour Street English: pseudo-archaisms like thou, hath, spake, twain, forsooth, howbeit that were already extinct in Murray's own day. The Riordan revision strips these out, removes divergences from the AG that add now-redundant context, with reference to Pope's verse where Murray's prose falls flat.
Cunliffe's Lexicon of the Homeric Dialect (1924) and 8,000+ Iliad line citations from Liddell-Scott-Jones serve double duty: as tap-to-lookup dictionaries with structured definitions, and as training data for Dragoman, a word alignment model fine-tuned on lexicon citation pairs rather than parallel text alone, reaching F1 0.80 on AG-EN and 0.61 on AG-MG (up from 0.65 and 0.48 with base UGARIT). Citation pairs teach Dragoman which AG words map to which English glosses, sharpening alignment accuracy on Homeric vocabulary.