This post is part of a series on my process and progress working on a capstone project for General Assembly’s Data Science Immersive course.

In the last post I began to flesh out the ideas I had for my DSIR capstone project. This project is the only one we really get to choose, and it becomes a bit of a passion project. I explained my backstory a bit and revealed what made these ideas so interesting to me. So here is a quick summary of the projects again:

  1. A tool/script/webapp etc. to automatically segment ancient Greek text according to the (big word incoming) morphophonemic method.
  2. A tool/script/webapp etc. to scan prosody and output the metrical units etc.
  3. Make an attempt at predicting authorship of texts based on…well that’s the data science part I guess.

The last post went into some detail on number 3, so today let’s focus on number 2.

According to Wikipedia, prosody is the is the “theory and practice of versification”. You may have learned it as “scanning” poetry in high-school etc. In English you generally use the stress of the words in a line of poetry to figure out the meter, which is not always so easy to do, especially when you run into strings of words with just one syllable. In ancient Greek, instead of word stresses you use the length of syllables. Using English syllables as an example:

mat is a short syllable

late is a long syllable

Seems simple enough. If only it were truly that easy. The following list is a sample — also from Wikipedia — of the rules and exceptions that apply to ancient Greek prosody.

Note: It is not important to actually understand or even read these, it’s just for shock value. However, the italic parts are relevant later in the post.

There are rules that determine the length of any given syllable. A syllable is said to be “long by nature” if it contains a long vowel or a diphthong:

η and ω are always long.

α, ι, and υ are long in some instances and short in others.

αι, αυ, ει, ευ, οι, ου, ηυ, υι, ᾳ, ῃ and ῳ are diphthongs.

A syllable is said to be “long by position” if the vowel precedes the consonants ζ /zd/, ξ /ks/ or ψ /ps/ or two other consonants. However, two consonants will not necessarily produce a long syllable if the combination is a plosive followed by a liquid or a nasal.

π, β, φ, τ, δ, θ, κ, γ and χ are plosives.

λ and ρ are liquids.

μ and ν are nasals.

The final syllable of a line, even if short by nature, is always considered long (“brevis in longo”).

γμ, γν, δμ and δν will always make a long syllable even when preceded by a short vowel.

A long vowel or diphthong preceding a vowel may be short. This is called “Epic Correption.”

Apologies for the language, but it’s accurate

Apologies for the language, but it’s accurate

These are general rules, there are even more rules depending on the meter of the text you are reading. You might be wondering when we’ll get to the data science part, well let’s get to it. If you noticed the “long by nature” and “long by position” you may have also noticed that in order to make it something done automatically, it requires a bunch of “if-else” statements. This of course has already been done before.

My goal would have been to try and train a neural network that learns these rules on it’s own. This would have involved a ton of data cleaning (not really a bad thing), and finding texts of already scanned lines proved to be quite a pain. To be honest this idea wasn’t of too much interest to me and my knowledge in Greek prosody is pretty minimal, you now know about as much as I do.

Stay tuned for my next post where I introduce my final topic choice and start to dig deep into the work I’ve put into it.

γνῶθι σεαυτόν
“Know thyself”
— Inscription at the Oracle at Delphi