The paper investigates the visualization of natural language's high-dimensional form in 3D. It begins by contrasting the widely circulated visual representations of DNA and viruses with the largely unexplored visual representation of language. The authors propose a multidisciplinary approach combining linguistic analysis with architectural and design expertise. They highlight the elementary structure of spoken language as revealed in transcripts, emphasizing pauses and intonation patterns that shape the organization of lines, comparable to verses in poetry. The research question centers on how aspects of this high-dimensional form can be rendered in 3D. The paper will trace the historical comparison between language grammar and geometry, focusing on evidentiality as a chosen grammatical feature for 3D modeling. The methodology will be detailed, showcasing the step-by-step process of creating a 3D digital image from language samples. Four prototypes based on diverse language samples (Kurdish, Tariana, Neo-Assyrian Akkadian, and American English) will be presented, alongside a discussion of the project's contribution to computational design and linguistic understanding.

The paper draws upon historical sources comparing grammar to geometry, citing Robert Kilwardby and Spinoza's influence on Whorf's concept of "a geometry of form principles" characteristic of each language. It emphasizes the abstract nature of grammar, similar to geometry's abstraction from concrete reality. The authors discuss Jakobson's work highlighting pronouns and evidentiality as features anchoring language to material reality. The existing literature on evidentiality, particularly its extensive documentation in Native American, Amazonian, and Australian Aboriginal languages, is reviewed. The research also draws on work concerning narrative patterns and texture, alongside the history of design, specifically spline-based software used in film and video animation. The historical roots of parametric design are explored, referencing Gaudí's use of weighted strings and the principle of inversion in architectural design. Finally, the paper connects its approach to classical antiquity's visualization of language as pattern-weaving and sewing, linking it to the modern concept of "text" and the figurative texture of language.

The 3D modeling process maps linguistic features onto a Cartesian coordinate system (x, y, and z axes). The x-axis represents the timeline of speech, the y-axis represents the number of syllables per line, and the z-axis represents the evidential weight of each utterance. The process involves manually formatting transcripts to indicate the relative weight of evidentials numerically. This weighted transcript is then inputted into an Excel spreadsheet, read by a Grasshopper script (integrated with Rhino 3D modeling tools) that counts syllables and identifies evidential weightings. The data is organized along the x, y, and z axes, creating a digital weave. The z-axis incorporates the evidential weight as virtual “weights of evidence” along digital splines, creating a 3D wireframe. The script allows for the import of numerous transcripts, rebuilding the geometry based on the input data. The software’s morphogenetic function then creates a smooth, woven surface from the wireframe, resulting in a 3D model. The authors describe the selection of language samples, detailing the coding of evidentials based on existing linguistic research for each language, including Kurdish, Tariana, Neo-Assyrian Akkadian, and American English. The process highlights the conversion of linguistic data into visual programming, with the creation of an algorithm that generates the 3D geometry based on various parameters. The resulting models are intended to be 3D printed objects.

The study presents four 3D-printed prototypes, each representing a different language: Kurdish, Tariana (Amazonian), Neo-Assyrian Akkadian, and American English. The Kurdish model shows regular undulations representing frequent evidentials and highlights showing reported speech. The Tariana model, reflecting obligatory evidentiality, features crests in almost every section. The Akkadian model exhibits a textured middle section with highlights showing reliance on reported information. The American English model, conversely, is characterized by smoother undulations reflecting the non-obligatory nature of evidentials in the language. These variations in the 3D models directly correspond to the grammatical differences in how each language handles evidentiality. The authors highlight the visual contrast between the Amazonian and English models, emphasizing the grammatical divergence between obligatory and non-obligatory evidential marking. The 3D models are presented as "reality-effects," showcasing the fusion of language and observable reality. The unique shape of each prototype represents the diversity of language and its interaction with reality. The authors argue that these visual representations offer a deeper understanding of evidential systems and linguistic relativity, particularly in languages where translation often fails to capture their nuances.

The findings demonstrate the feasibility of 3D modeling as a means to visualize the high-dimensional structure of natural language and, specifically, evidentiality. The distinct shapes of the four prototypes highlight the linguistic relativity of evidentiality and demonstrate the potential of this method for revealing subtle grammatical features that are often missed in traditional linguistic analyses. The visual nature of the 3D models provides a more intuitive understanding of evidentiality, surpassing the limitations of textual descriptions. The study’s success in visualizing language’s interaction with reality suggests a new perspective on language documentation, especially for endangered languages where traditional methods may fail to fully capture the complexity of grammatical features. The research contributes to both computational design and linguistic anthropology, opening new avenues for investigating the relationship between language structure and perception.

The paper successfully models aspects of natural language in 3D, using evidentiality as a key feature. The four prototypes demonstrate the unique characteristics of different languages and the power of visualization in revealing nuances lost in translation. The use of parametric design offers a novel approach to form-finding in linguistics, extending the legacy of parametric design to a new domain. Further research could involve larger datasets, moving images, and exploring additional grammatical features beyond evidentiality.

The study focuses on relatively short language samples, limiting the generalizability of the findings to larger datasets. The 3D printing process faced technical challenges, with many models rejected due to complexity. The selection of specific language samples reflects the researchers’ expertise and access to data, potentially biasing the results. Future research could investigate the use of alternative modeling techniques and printing methods to overcome these limitations.