Getting a bit closer to having a typed syntactic parser for Ancient Greek (given a lexicon). Currently, I can infer MERGE types (direction and application vs. composition), but it still requires some hints as to constituency in cases where the structure is not all nesting in one direction.
So the following string gets parsed perfectly without any help:
τὴν Εὐρυτείαν οἶσθα δῆτα παρθένον
But something like the following needs some brackets to help the parser understand what's up:
ναυμαχία [ γίγνεται [{∅-adv} ἐπ' {∅-det.} Αἰγίνῃ] {∅-det} μεγάλη]
I hope to make a thingy soon that will let it guess this structure properly, so that it will be able to parse this string:
ναυμαχία γίγνεται {∅-adv} ἐπ' {∅-det.} Αἰγίνῃ] {∅-det} μεγάλη
Of course, for anything that it has parsed, it can then “evaluate” it into a constituency-based tree, as follows:
[ {∅-adv} [ ἐπ' [ {∅-det} [ Αἰγίνῃ ] ] ] [ γίγνεται [ {∅-det} [ μεγάλη [ ναυμαχία ] ] ] ]
The work is here: https://github.com/jonsterling/Dingle-Dangle/blob/master/haskell/Syntax.hs.