________________________________________________________________________

Somatic dance speech technologization in four steps
The technologization of somatic dance speech captures and analyzes the verbal expressions that arise during visualization (Phase II) and reflection (Phase IV) in Nervous Systems. These expressions are recorded as audio files, coded and organized within a database that identifies recurring patterns and tendencies in what dancers and hosts articulate. The database then feeds SOMAi beta, a dance-inspiring software that automates support and feedback for dancers during movement exploration (Phase III). The central challenge designing SOMAi beta is to preserve dance speech’s co-creative, meaning-making, and socializing role while keeping the system and audio samples open to variation, interpretation, and ongoing transformation in improvisation.

The methodology proceeds in four steps: (1) data acquisition; (2) taxonomy and (de)coding; (3) map and interface design; and (4) software development and testing. The outcome is an open, interactive map available online and in the Dance Data Object I. This section of the dissertation functions as a choreographic object, including the database inventory, the interface design, conceptual categories, indexed key terms, and preliminary linguistic analysis.

1. data acquisition
The process begins by participating in somatic dance practices, either as host or as dancer. Before each session, I briefly explain my double role and request permission to audio-record the dance speech for research purposes, assuring participants that all recordings will be anonymized and could not be traced back to specific settings. During the practice, the intervention is minimal: I used a Zoom H5 recorded, which captures high-quality audio and is compact enough not to distract from the practice. At the end sessions that I did not host, I invited the host for a short conversation to say thanks and address any questions; these conversations were not incorporated into the dissertation.

I transcribed recordings two to three weeks after each practice. In ongoing workshops, this delay helped me to stay engaged with the group and avoid prematurely formalizing or intellectualizing the work and disrupt its intuitive, embodied nature. It was also practical given the pace of the acquisition period. Many of the sessions took place during Impulstanz festival, where I taught and participated in multiple courses daily. The delay had an interesting effect on memory: listening back to the visualization exercises easily brought me back to specific moment in practice, a recall activated by my muscle memory. By contrast, relistening to reflective speech was less affective and made the temporal gap more tangible. I return to these observations in Chapter 7, where I discuss how they informed strategies for shared exchange retrospection within Nervous Systems.

2. taxonomy and (de)coding
I began by cutting up and normalizing the transcripts using the speech protocol outlined in Phase IV (Chapter 3). All utterances were rewriting in first person singular present tense. Each guiding statement was then paired with a reflective counterpart (and vice versa), producing matched couples of guiding statements and reflective questions. These pairs were listed into spreadsheets: questions received odd numbers; their paired statements, the following even numbers. I re-recorded every sample in this double form myself to anonymize source and standardize audio of the total 2098 dance speech expressions.

In a macro-analysis of the database, I identified eighty recurring key terms. These terms were cross-referenced with established 20th-century Western movement and dance frameworks (e.g., Laban Movement Analysis, Body-Mind Centering, Bartenieff Fundamentals) which allowed fine-tuning the analysis into a single organization system for numerical coding and mapping the database in branches roughly representing the different analytical movement approaches.

Each key term was assigned a numerical code. In doing so, each sentence received a unique index in which a chain of codes represents the combination of different key term that is associated with that dance speech expression. This index enabled us to organize categorical libraries per key terms, an essential step for the later software operations that allow the system to combine data, chaining sentences back together into a meaningful narrative.

3. map and interface design
Mapping unfolded over an extended period. We tested several visual organization models to represent relationships between key term and the movement analysis approaches used to organize them. We settled on a collapsable map with adjustable levels of detail. This design allowed us to organize and present the data in a structure that is equally accessible to users without somatic or movement analysis training and seasoned dancers who prefer analytical depth. The map centers on a circle with the SOMAi logo around which four nodes present the main categories: senses, body, motion, physics. A fifth main category captures interrelations across the main categories; we hide it in the interface to avoid unnecessary complexity in the starting setup. The four nodes branch out into their respective subcategories in opposing directions. In total, the system comprises four main categories, twenty-two first-order subcategories, and fifty-four second-order subcategories, each with its own key term library. We assigned a color to each branch and created short animations to accentuate the appearance and disappearance of categories. The detailed taxonomy and the small-scale structural and linguistic analyses of the datasets are presented in Dance Data Object I.

4. software development and testing
SOMAi beta is a tool for movement exploration. This semi-controlled system encourages bodily and situational awareness by generating dance speech within the improvisation; it functions as a digital co-host in practice. Its engine recombines pre-recorded speech samples using controlled randomness (probabilistic logic). By assembling samples that share the same codes in their unique index, it produces a coherent stream of suggestions, guidance, and question for the dancer.

The graphical interface (see Dance Data Object I) displays categories as unfolding nodes and branches, each corresponding to a key term library. Clicking a term activates that library, clicking again deactivates it. Users can navigate the map by interest or need through two operational modes: random and selective. Random mode supports open improvisation, sequencing material from all branches into a dance speech narrative. Selective mode lets dancers prioritize chosen key terms; for example, activating breathing out, fast movements, and lower limbs restricts the output to those libraries. Users can further refine the timing and dynamics via sliders: set the balance between questions and statements, decide the duration of the narrative, the speech tempo, and vocal timbre along two axes (human and computerized, high and low pitch). Pressing the central button start and resets the score generator.

In the next chapter, I detail how we this methodology, the resulting database, and SOMAi beta are integrated into the Nervous Systems praxis. In workshops on festivals and in courses and in educational settings, I have implemented elements from Project SOMAi. Additionally, together with Meurisse, I am exploring to repurpose the software architecture to pair dance speech with other data and media, opening intermedial processes and work with interdisciplinary groups as start exploring in the closing performance Anatomy of New Organs.
________________________________________________________________________