Highlighting significance in doctoral thesis conclusions: A SciBERT-based analysis of adjectives of importance across disciplines

The conclusion chapter of a doctoral thesis is important to the thesis's overall reception and potential impact. However, the linguistic resources used to highlight research significance in conclusion chapters remain underexplored. This study employed SciBERT, a language model pre-trained on scientific texts, to computationally analyse adjectives of importance in 150 doctoral theses across the disciplines of applied linguistics, physics, and psychology. Using UMAP dimensionality reduction, k-means clustering, and permutation tests, this study examined the overall distribution, semantic clustering, disciplinary variation, and syntactic functions of these adjectives. Results revealed that important and significant dominated across disciplines, with a two-cluster ( k = 2) semantic structure emerging as optimal. Eleven adjectives showed statistically significant disciplinary differences, with physics theses using most of them less frequently. Syntactic analysis revealed a dominant use of attributive functions. The findings demonstrate how doctoral students strategically deploy adjectives of importance in their thesis conclusions and offer insights for discipline-specific instruction in thesis writing. The computational analysis also extends methodological applications in academic writing research.