High Performance Sound Technologies for Access and Scholarship, is a DH project headed by Prof. Tanya Clement. The projects is based out of the Univ. of Austin Texas’ School of Information and works with the Illinois Informatics Institute located at the University of Illinois, which focuses on the STEM related bioinformatic research. A simple reading of the project’s title, entails inputting audio files into computer softwares and extracting analysis for future scholarship and preservation. Her powerpoint presentation was titled “Recital, Machine Learning and Visualization with Sound”. At the end of the presentation I found that Tanya Clement covered some subtopics better than others.
Following an essential DH trademark, HIPSTAS has been funded by NEH grants after the release Library of Congress’s National Board of Sound Recording 2010 article “A National Legacy at Risk in the Digital Age”. Acting as a platform, the article targets hours of historical audio at hidden in various archival collections. An extensive survey concluded the volume of historical audio records at risk for destruction through deterioration and four major obstacles to their access 1) conservation and preservation formatting 2) barriers to public access 3) advance professional development 4) outdated laws that impede preservation and access (LoC National Board of Sound Recording). Physically bulky and costly to digitize or transcribe, audio collections often reside in archive holdings unnoticed by preoccupied archivists beyond the knowledge of the public and researchers. Additionally archival tradition of paper based material, keeps many such audio collections neglected and beyond the immediate scope of any More Product Less Process initiative.
Following the collaborative aspect of many DH projects, Tanya opened her presentation citing her outsider identity and lack of sound and music knowledge. Ultimately it was but desire to produce humanist based research by filling an information gap. Her interest lies in scholarly information infrastructure development and collaborating to literally fill information gaps. Such matters horizontal efforts which entails networking and touring non-academic spaces to bring in outside talent, opinions, and funds.
ARLO (Adaptive Recognition with Layered Optimization) is the audio analyzing digital tool behind HIPSTAS, originally developed and used by ornithologists to classify bird chirping patterns, especially in aurally dense settings like a rainforest canopy. It is multifaceted open source program that analyzes or ‘reads’ ten second audio clips and outputs visualizations based on sound wave frequencies. The visualizations are based on a color spectrum from white (highest pitch) to black (no pitch at all), which can be further broken and analyzed. These analyses are the paratext with their own vitality that add to the richness of the original information being analyzed (Ramsey).
Digital humanists have noted the difficulty of searching for tangible information in audio format (Fujinaga and Weiss). These conclusion are based on audio’s inherent technical opacity. An ephemeral data source, audio has to be captured by the right parties equipped with the proper tools. Instances of intrepid musicologists like the Lomax family come to mind, who travelled across America with a massive phonograph recorder in their car trunk.
ARLO can extract through clustering an array of subtle data set or patterns found in audio files based on pitch, timbre, and mood. These extractions are evidenced through methods following the scientific principles of hypothesis, testing, analysis, and conclusion. Homonyms, homophones, synonyms, and mondegreens can be compared side by side using a variety of variables. Familiar and consistent blocks of audio such as speeches or poems spoken years apart can also be juxtaposed and compared.
ARLO the machine and HIPSTAS the operational body converge for a factually credible and fruitful digital humanities project mining established audio and oral history collections from Native American Ojibway and the Oil Merchants at Univ. of Texas Austin. Its use by oral historian and anthropologist are an obvious collaboration and historians and librarians hope to use ARLO’s digital analysis to make conclusive decisions about the quality of large unprocessed audio collection. Its use by the Ojibway from Minnesota and other indigenous tribes focuses on the project’s preservation capability and its bring light to disappearing anthropological traditions such as folk legends, story telling, riddles, and initiation rites. This collaboration boosts the project’s credibility and places it within an established national tradition of recording folk music and related audio.
HIPSTAS’s secures it credential from its working relationship with non-traditional sound and music fields, one of them being UPenn’s PennSound collection cored around audio versions of American poetry. The web based poetry and spoken word portal has been mined with HIPSTAS by enterprising English scholars willing to lend credence to the veracity of spoken word and poetry. While oral historian will still possibly trudge every aural nook and cranny, non-sound archivist will find its hearings and merits about an audio collection helpful regarding accessioning and processing. This in turns adds to an easy use digital toolbox for less technical savvy archivists to deal with audio collections adequately.
Bibliography
Burdick, A., Drucker, J., Lunenfeld, P., Presner, T., & Schnapp, J. (2012). Humanities to digital humanities. (pp. 3-26). Cambridge, Massachusetts: MIT Press.
Fujinaga, I., & Weiss, S. F. (2004). Music. In S. Schreibman, R. Siemens & J. Unsworth (Eds.), A companion to digital humanities (). Oxford: University of Oxford.
Ramsey, S. (2011). Reading machines: Toward an algorithmic criticism. Chicago: University of Illinois Press.