Sappho in Data

@yum25
44196941631255817982281271375595301562112601411682071241501910310521104111576514887158133181061181558516588633232649401311386813256411511412051527014464931404765416610115673122161167173109741081161642584163192125911915317717969170172174176180181183184185186188189190191431105013039135127146187864633147761693612518271131609291078361145341547842149328011737126143151142121102481345392100128136152162678212991413515945124838123157175

Categorizing Sappho

The literary corpus of Sappho is rife with references to mortals and gods alike - Aphrodite, Gongyla, Atthis, and brimming with descriptions of beauty and nature.

The strong thematic similarity between poems, both in who they invoke and what they describe makes Sappho's fragments an especially interesting subject to place in distinct categories. I wondered: what if I tried to group these poems using a machine learning algorithm?

Through measuring the number of words in common between fragments and a little bit of math, I created a data visualization where each distinct grouping is repesented with a different color compared to those around it.

These fragments are translated by Anne Carson, in her collection: If Not, Winter.

Scroll to start exploring! ↓

Each circle represents a fragment.

The number on each circle represents the fragment number.

The size of the circle represents the approximate length of the fragment. That means if a fragment is longer, its circle will be bigger, and if a fragment is shorter, its circle will be smaller.

It can be a bit difficult to see some of the shorter fragments. To read the fragment number, hover over the smaller fragments to enlarge them!

First, let's look at some of the largest and smallest groupings.

The algorithm used to group the fragments generally associated fragments of similar lengths together.

This makes sense - larger fragments have more "gravity" - the longer they are, the more words they have, and the more words there are, the more likely the algorithm will find word similarities between fragments.

Notably, some of Sappho's most well known and frequently translated fragments are grouped together, such as the Ode to Aphrodite (Fragment 1) , Fragment 16, and Fragment 31.

The smallest cluster is made up of 3 short fragments.

Fragment 123
just now goldsandaled Dawn

Fragment 157
lady Dawn

Fragment 175
dawn

All mentioning dawn. Interesting! Let's see what other fragments have the word "dawn" in them...

dawn

Even though there's a strong correlation with the smallest cluster, it looks like the word dawn is mentioned elsewhere as well.

Let's take a look at some other words! They might reveal interesting patterns in how often words appear among different fragments, and whether it is a strong indicator of where each fragment gets categorized.

Aphrodite, Kypris, Kypros

Eros

gold, golden, goldhaired, goldsandaled

moon, silver, silvery

That was interesting!

While some categories, such as Aphrodite, Kypris, Kypros and gold, golden, goldhaired, goldsandaled appeared all across the map, other words such as moon, silver, silvery were more closely associated within the same few groupings.

Most interesting of all, all instances of the word Eros showed up in disparate categories.

Perhaps instead of being a subject of Sappho's poems, Eros is a constant.

It's time to hand over the exploration to you! Thanks for reading!

Search for any word to fade in and out fragments that have those words. To search for multiple words, separate each word with a comma.

Like what you see?
Check out my other work! :)