Initial remarks for the second gathering of the High Performance Sound Technologies for Access and Scholarship institute at the University of Texas at Austin, May 27-28, 2014.

===

Given my position in today’s proceedings and the specificity of the papers that follow, I thought I might transition from Tanya to begin the morning’s papers with some general remarks on sound recordings, poetics and critical practice directed specifically toward ARLO and our work as a research cluster.

In these remarks, I hope to do a few things: to reflect on this past year of research activity and the confluence of this group of people who are all pursuing various aspects of recorded sound; to begin to frame the discussions that follow in terms of what kinds of research we might be able to produce through this confluence, individually and collaboratively; and, finally, to advocate for a kind of grey literature, a mode of technical reports that might facilitate a discussion of where we can go from here and what we can produce through our researches.

(In mentioning technical reports in the humanities, I’d like to acknowledge Nick Montfort’s article on specifically this topic that was published in the first issue of Amodern.)

My remarks focus in on three specific questions: What are we trying to produce here as researchers? What have we generated in the past year that might provide direction for future research? And how can we organize the communications between us so as to realize particular trajectories of the research with our various repositories of sound and ARLO?

In regard to the first question – what are we trying to produce here as researchers? – in our experiments with tagging, clustering, and our individual tests, there have been a number of occasions in which I’ve had to remind myself that we don’t exactly know what we’re attempting to produce through experimentation with ARLO. There’s not one specific end to work toward.

I see this is a productive situation to be in. It means that researches can be directed toward a number of outcomes that take place on various levels or trajectories: relating to archival or technological aspects of the sound recordings, to the more theoretical dimensions of working with sound and performance, or to the development of linguistic or pedagogical tools.

I wanted to emphasize this simple point because – at least in the PennSound gang that’s been in regular dialogue over the last year – sometimes I think we might overlook certain small but important things we’ve generated in our experiments and dialogues while in the process of trying to work out some kind of major breakthrough.

So, while in the process of running experiments toward that breakthrough – say, in which we teach ARLO through machine learning to maintain our various repositories (ie. being able to notify us of redundant files, or to track down tampered recordings through thousands of files) or in which one or some of us model(s) a well-tested experiment with ARLO that changes how we approach phonotextuality – let’s not overlook the steps along the way that might produce a critical tool or technological tool (or both) for working with sound recordings.

Which is to say, let’s keep in mind the second question – what have we generated in the past year that might provide direction for future research? Here, I have in mind a few examples that have stuck out in the last year’s experiments and dialogues:

Visualizing Sound. There are many aspects or angles of this discussion that come to mind. To begin with the most basic, both in our meeting last year and the in the meetings since, one thing that we continue to discuss are the various surprises we have working with visualized sound. Often, say, in working with tagging, we catch various aspects of a recording that we never would have heard without that kind of close listening. Or, especially in the clustering experiments, there have been a number of times we can’t quite hear some aspect of a recording until we see it visualized or its difficult to hear an affinity between two recordings until we see them visualized. This activity has stemmed numerous discussions, many of which could be exceptionally productive if brought into a classroom setting.

Here, I’m imagining a simple tool in which people use ARLO in its most basic way to visualize excerpts of sound recordings. In this instance, let’s use a PennSound recording as the example. Alongside of working with the text of a poem, students would be able to, say, visualize two different readings of the poem. With these visualizations they could then begin to talk about rhythms in the read poem or identify patterns in it. Or the difference between the two recordings of the poem in relation to the text. They could discuss the trace of the technology that is present in the sound recording and move toward a discussion about media. What I’m envisioning here is the production of tools to decenter the emphasis on the page-text in pedagogical settings in an attempt to advance engagement with sound and performance and technology.

The Gender Button. This issue turned into an interesting and somewhat divisive discussion that took place during our initial tagging experiment of whether or not one should assign a gender to the speaker heard. The argument to assign a gender was to assist with a machine learning in which ARLO would be able to pick up on specific tonal ranges in which, for example, a “high tone” and “male” tagged clip would not get confused with a “moderate tone” and “female” tagged clip (as they might be visualized similarly); the argument not to assign gender to a voice having to do with the problematic conception of gender implied by such a tagging, the problematic binary that is maintained in such a conception, as well as the act itself of assigning gender to a voice, which many would argue is a enactment violent or violating form of judgment.

Here, I can imagine the discussions that were carried out in regard to this tagging could play an important role in an article on ARLO experimentation, one that would consider the gender politics of the voice and sounds in general, an issue that to the best of my knowledge is understudied or has not been sufficiently addressed. (The only example I can think of off the top of my head is Tara Rodgers’ Pink Noise.) This topic was one that has been addressed from different positions and with great nuance, and yet it’s one that has sort of slipped through the cracks in working toward other goals, which I think is unfortunate. It also leaves something to be said about documentation of our activities and the report-based mode of internal communications that I will come to in a moment.

Poetry / Not Poetry. This, like “the gender button,” was part of the tagging project that most of the PennSound gang carried out. Like the proposed gender button, it operated on a kind of binary logic in which one listens to a two to four second clip from PennSound and decide whether the sounds heard are Poetry or are Not Poetry. The purpose behind such tagging was to see if it might be possible to teach ARLO to recognize what is Poetry and what is Not Poetry. The possible results of such machine learning aside, I found endeavor of the tagging itself to be an interesting project, particularly when one considers that many of the poets included in PennSound are ones who are often trying to always work beyond the more normalized or historical confines/classifications of what is and what isn’t poetry – formally, textually, rhetorically.

Here, again, I can envision a simple pedagogical tool in which students listen to a clip of sound – visualized or not – and make the decision to label that clip as Poetry or Not Poetry. Mixed in with more normalized metered verse recordings could be recordings, say, from talk poets David Antin or Robert Ashley, or from the ritualized performances of Cecilia Vicuna or bill bissett, or from Laurie Anderson’s poem-opera United States. The results from the students’ tagging could provide a truly compelling way for them to consider poetry and to do so based in a way centered upon its sounding.

So, to return to the bigger picture here, I’m arguing that while trying to make that individual or collective research breakthrough we keep in mind the discrete points we cross along the way and harness those points to develop either written research or to produce digital humanities tools that can have a direct pedagogical impact in getting students to experiment with sound recordings. I’ve used examples from discussions of poetry sound recordings, but see this working just as well working with sound recordings for Indigenous languages and folklore recordings and others.

Now, all of this leads to the final point I want to make, which is in regard to the inscription of research activity as we carry out individual and collective projects. I simply want to suggest that we begin to utilize a mode of publishing amongst us as a way to work toward more official research production and publication that goes beyond this group.

Above, I took up the term technical report, which I think the example coming out of science-based research is useful in our own research setting. For our own purposes, this mode of publishing can be something as simple as maintaining a site – a simple wordpress site would work fine – where individually we begin to publish every couple of weeks or once a month short reports that relay between us what we are thinking about and what we are doing. (It would be an exceptional start to begin with the talks/notes we’re presenting today.) What I’m imagining here is a central document of our overall research activities: the short reports, which I’d say are central to this publishing mode, but also notes on experiments, minutes from the monthly meetings (as well as the sidebar chats that happen during those meetings) and the occasional updates from Tanya, David and others.

With these working notes or short reports we can begin to find concrete points to engage and support one another’s research activities, develop possible collaborations, and produce citable texts that can provide a base for more official research articles or the production of pedagogical tools.

I know Tanya has offered access to the main HiPSTAS site and Al has encouraged us to publish on Jacket2. I think that both of these sites are exceptionally helpful for disseminating the more official research that extends beyond this specific community. The mode of publication I’m suggesting, though, is something a little less polished and more internally directed toward the HiPSTAS community: technical reports so that we can begin to build off of one another and find points of convergence to develop ideas in a kind of laboratory-styled model of research production.

(Here, I’d like to acknowledge Darren Wershler who has introduced me to this style of research production in the humanities, an approach he’s advocated and explored in a number of ways – the site I’m writing this on, AmpLab, being just one realization.)

In remembering that we don’t specifically know what we’re trying to produce through experimentation with ARLO, and in realizing that their can be a number of outcomes that take place on various levels or trajectories of the research, and, finally, in noting the special circumstances of all of us coming together through this project and each one of us bringing a different set of skills to the table that in a more highly collaborative setting could produce something beyond what we may have originally envisioned in working with this project, this report-centered mode of publishing will help direct the various trajectories of research toward more official articles or toward the realization of tools that could have a profound effect on the way that the language arts are conducted.

===

Cover Image: ARLO Visualization of lines from N.H. Pritchard’s “Gyre’s Galax” with accompanying text.