We have been working with the Department of English & Comparative Literature on “Linked Reading,” which uses audio machine learning techniques to allow researchers to directly query, analyze, and visualize data from multiple independent datasets, including UC’s Elliston Poetry Archive. Since 2010, UC Libraries and the English Department have collaborated on the Elliston Project, an audio archive of over 700 recordings of poetry or poetry-related content. The recordings span seven decades and include over 450 poets, including Wendell Berry, Robert Frost, Allen Ginsberg, Louise Glück, and a host of others. Alan Liu, a pioneer of digital humanities in literary studies, considers the Elliston project to be a “world-class poetry audio archive,” which has the potential to “alter the dominant understandings of a ‘digital archive’ developed for textual materials.”
To accomplish this examination of audio poetry archives, we have entered into partnership with the University of Arizona’s voca poetry center to create a linked data infrastructure and sound analysis platform that allows us to query and compare specific poetic lines and words in multiple versions of a single poem housed in the two archives. Using a method that we call “Linked Reading,” we are investigating the reading styles of poetic schools and the formal dimensions of spoken poetry that differ from the printed text, by using semantic analysis of the printed text, coupled with sonic analysis of the audio archives. This approach, gathering smaller datasets of creative materials into a linked network, allows our team to leverage the local strengths in creative writing and electronic media to facilitate a new range of projects between poets, digital humanists, and scholars of sound. Our pilot research project compares multiple readings of authors represented across the two archives to examine how poets recite meter, how they use enjambment, punctuation, and breath in their performance, and how prefatory comments and jokes delivered before and after a poem add context to the formal dimensions of the lyric, precisely as marginalia create a supplementary layer of meaning in a manuscript or printed text. It is a basic tenet of poetry instruction that the poem on the page is but a score to be performed; and yet, poetry scholarship is in virtually every instance a study of the printed poem. The project illustrates a potential of reshaping this well-studied topic by ‘reading’ the sonic features of poetic form across a large number of poetry recordings in multiple linked audio archives.