There are those who say that Paradise Lost is the best poem ever written for the English language. There are certainly a lot of poems out there, many of which are great.
However, such claims are absolutely correct. Nowhere else will you find a man boldly explore as many burning questions of humanity, existence, and morality, and answer them with such raw passion and emotion. The book is an amazing, transformative experience.
The idea for this short text analysis exercise came about while I was taking a break during a busy Wednesday evening. Just like my previous graph, it came with the thought “Wouldn’t it be funny if something significant were to happen…”
In this instance, it was verbs. Specifically, all of the verbs that each character uses in Paradise Lost, ranked by frequency.
Like most projects, my first step was to think of a rough outline for how this could be done. Regardless of the tools I end up using, generating a list of methods towards finding a solution helps me be clear about my goals. I wanted to:
- Locate each named character (I just used SparkNotes for this)
- Find each mention in the book
- Search within 5 ish words of the character for verbs
- Rank by frequency
This was something I could have done myself, given enough time and patience. However, one of the nice things about being in the AI era (and having an assignment where we are legally allowed to use it) is that decently thought out ideas can be implemented very, very quickly. So, I told Gemini my plan and it gave me a pretty solid draft.
What distinguishes vibe coding from AI-assisted development is the level of human contribution. Though Gemini did most of the coding, I gave it clear instructions on how to make it. I wanted the graph to be made using plotly, and suitable to be displayed on the web and suggested the regular expression that it use for sifting through words. I eventually realized that stopwords had to be added (for example “and”) to make the list meaningful. Near the finish, I realized that some character names (like “god” or “eve”) were apparently verbs, and so I manually edited those out.
Here is the result:
I was impressed by the results. I had a silly idea, and was able to execute it in an evening. There are, of course, many problems with AI, but the ability to perform complicated operations en masse is one of the positives.
The data gives a couple of interesting musings. God, for example, does a lot of making, creating and having, which checks out because, well, it is God. Adam and Eve both “reply” a lot, because they are constantly learning and reacting to the world. Satan sits on his throne quite a bit, and Death sins quite a bit. I find it mildly funny that the Son (Jesus) exists largely to “thus spake”, which, looking back, also makes sense, because John Milton often loaded a lot of flowery speeches onto the Son.
As for the assignment prompt, there are many different ways we can go about analyzing text:
- Which words appear the most?
- What subjects or ideas occur?
- What kinds of patterns emerge over time?
- Why kinds of patterns don’t emerge over time, and why?
The awesome thing about having all of these tools to play with is that answering these questions can be much less work, fun even.