Modeling variation and change in scientific English over 300+ years from a communicative perspective

One major aspect in our research is to understand the role of variation in helping to modulate the information content of linguistic units, achieving optimization effects for efficient communication within the realm of scientific writing. By situating our inquiry in this controlled environment, we gain nuanced insights into how language variation and change contribute to the dynamism of information exchange in scholarly discourse.

I will present how we can detect and analyze variation and change in language use with data-driven methods that apply information-theoretic concepts without a preselection of theoretically motivated linguistic features. Traditional linguistic research often requires the preselection of specific linguistic features based on theoretical motivations. However, our method bypasses this by focusing on inherent patterns in the data itself. This enables us to uncover latent linguistic variations and changes that might not be immediately evident or might even be overlooked in traditional analyses.

While the primary focus of our study revolves around linguistic features, the methodologies we employ hold a broader applicability. The data-driven and information-theoretic methods we utilize can be extended to analyze other types of features, revealing change and variation in different contexts. This versatility underscores the potential of this approach, not just in linguistic studies but in any domain where patterns of change and evolution are of interest and where probabilities of these changes can be generated.

A notable facet of this methodology is its ability to capture the inherent asymmetry in linguistic variation and change. In many linguistic processes, the directionality matters. This becomes especially pertinent when studying knowledge evolution. By capturing this asymmetric nature, our models can offer insights into how knowledge diffuses across interdisciplinary settings. For instance, the flow of knowledge between two closely related disciplines might be smoother than between two vastly different ones. This asymmetry, which is integral in a communicative context, can shed light on the genesis and evolution of disciplines, potentially pinpointing moments when the divergence in knowledge led to the birth of a new discipline.

While our methods help us identify significant linguistic shifts, qualitative methods are crucial to understand the reasons behind these shifts, which besides communicative aims could also be related to extra-linguistic factors (e.g. change in language use during the chemical revolution). Thus, while our methods can tell us "what" has changed, they do not tell us necessarily "why". In accordance with communicative accounts, we are interested in tracing effects of variation that modulate the information content transmitted leading to optimization effects. For this we apply additional information-theoretic notions such as entropy within paradigms and surprisal as a measure of predictability in context. On top, traditional historical research would provide the contextual depth to understand the reasons behind the linguistic changes we trace.