Wow, that’s a neat idea. I’d start by tokenizing the verse into its metrical units, then build a finite‑state machine that recognises enjambments and caesuras. Next, feed those markers into a recursive descent parser that looks for semantic shifts—like when a metaphor turns into a literal image. It’s like hunting for hidden code in a sonnet, but instead of a compiler I’d use a corpus of annotated poems as the reference set. The trick is to let the algorithm learn the rhythm‑meaning relationship, then you can query “what hidden pattern does this stanza hide?” and it will spit out a mapping of form to metaphor. Pretty much a linguistic de‑compiler, if you will.

Nice flow—coffee on standby, I’m all in. Let me just toss a quick note into the annotation pile and we’ll see what hidden code the stanza whispers.