cuny-cl / latin_scansion Goto Github PK

Should run the scansion test, if that's possible (see https://stackoverflow.com/questions/58849994/running-tests-at-circleci-in-conda-environment for how to get Conda running in CircleCI).

Thrax 1.3.7 has a trivial but fatal registration bug with StringFile. I will make an upstream release of a new version, get it into Conda-Forge, and then migrate this library to it. This should clear the CI failure. I am assigning myself.

Pynini 2.1.5 was just released and we may want to update. It doesn't have any major API or performance changes so I think we can just edit the environment.yml and/or requirements.txt files.

A script should be used to confirm that a Document textproto file is parseable. (As a side effect it could optionally reserialize the data in question to ensure it's as canonical as possible.)

This will allow us to manually add comments to the files (e.g., a note to the effect that some verse requires hypermetric), but ensure we don't break anything adding these annotations.

ūnius ob noxam et furiās Ajācis Oīleī?

Does my scansion match up with what you would come up with?

If so, I think furiās is [fur.jas] and Oīleī is [oj.lej]. Then essentially I would allow long i's to glide (in addition to the short i's), as part of synezisis?

My colleague Richard (politely) requests that we consider also supporting hendacasyllables, like some of Catullus's poems.

Technically this is not impossible but would be slightly annoying in that these are not traditionally described as having foot structure, just verse structure. Probably easiest to have a separate pipeline but reuse the lower-level (below the foot) grammar definitions, and then have a radio button to select the type of verse.

Just putting this here for later ;)

Currently the protobuf output by the Python scansion engine does not emit information about syllable boundaries, syllable weight, or foot structure. Only a slight redesign should be necessary to preserve this information and to store it in the proto.

n/t

If #94 is accepted, we will be checking in textprotos into the repo. We should perhaps add a test to the CI system that automatically runs latin_validate over any and all textprotos checked in to make sure that new PRs don't introduce syntax errors in them.

I'm reading Cser's The Phonology of Classical Latin and here are some things he claims we might want to support (or enforce)

1. s followed by a voiceless obstruent is always heterosyllabic: i.e., they must be split, there are no sC onsets word-internally.
2. s can resyllabify leftward, making a word ending in a short monophthong heavy, to fit the meter

In compound words with prepositions such as ad, ab, sub, dē, in, etc., I think the preposition should be treated as one syllable (but not always). As it is now, the meter grammar attaches the last consonant of the preposition as the onset of the next syllable. The issue would be to know in which cases the prefix in the compound word (if the subsequent syllable begins with a vowel) should be treated as one syllable, or whether the prefix should cause a split.

Example:
"ex numerō subit; ac magnō tellūris amōre"
subit is [su.bit] (i.e. the "b" is the onset)

"Ūnus abest, mediō in flūctū quem vīdimus ipsī"
abest is [ab.estː] (i.e. the "b" is the coda)