The distributional properties of long nominal compounds in scientific articles: an investigation based on the uniform information density hypothesis

Nominal compounds are a structure commonly used in scientific texts. Despite their commonality, very little is known about how they are distributed in scientific articles. Based on the Uniform Information Density hypothesis, which states that speakers communicate information at a constant rate, avoiding peaks and troughs of information transmission, we predict that nominal compounds should cluster toward the end of scientific texts, be preceded by supporting text that facilitates their understanding, and be repeated often after their first use. In this paper, we examine these predictions through a quantitative and a qualitative analysis of a corpus of scientific papers from the fields of Biology, Economics and Linguistics. While our investigation did not reveal definitive findings for the first and third predictions above, it did produce supporting evidence in favor of our second prediction, thus advancing our understanding of NC use and the choices speakers make when transmitting information.