My interest in the HiPSTAS project is split between the critical and the practical. I intend, first of all, to examine the messy prosody of poetry in performance. For the sake of statistical comprehensibility, I plan to analyze the distribution of amplitude peaks over time for a collection of poets’ readings, largely disregarding information on frequency and harmonic structure. My research question, in short, is how we can reconcile our intuitive sense of poetic rhythm with the unruly patterns of recorded speech.

What is the best approach for visualizing relative reading speed over the course of a performance? How do such variations map to tonal shifts in the work? How else can we describe the complicatedness/entropy of a speech pattern (i.e. degree of deviation from perfect periodicity)? Can we identify significant prosodic isomorphisms across a poet’s oeuvre? Can we isolate such parallels among poets with similar styles? Is it possible to distinguish between poetry performance and non-poetic speech for a given writer?

On the practical side, I’m interested in applying such distant reading strategies to archive management and curation. We at the PennSound poetry archive spend a significant amount of time segmenting and tagging newly digitized recordings in order to provide ready access to individual poems. Is it possible to automate some portion of our workflow? Furthermore, how might we use ARLO’s search functionality to assist in curating material for PennSound’s podcasts and online radio stream?

The above questions are tentative and incomplete, and I’m beginning this project with a healthy dose of skepticism regarding the value of broad claims in computational humanities research. Still, I look forward to comparing notes and fleshing out my ideas, both at our workshop in Austin and over the coming year.