Linguistic positivity in soft and hard disciplines: temporal dynamics, disciplinary variation, and the relationship with research impact

Abstract

Previous studies have investigated the use of positive/negative language in academic discourse, and have found a tendency toward using more positive language in academic writing. However, little is known about whether the features and dynamics of linguistic positivity vary across disciplines. In addition, the relationship between linguistic positivity and research impact deserves further evaluation. To address these issues, the present study investigated linguistic positivity in academic writing from a cross-disciplinary perspective. Based on a 111-million-word corpus of research article abstracts collected from the Web of Science, the study examined the diachronic trends of positive/negative language in eight academic disciplines, and explored the relationship between linguistic positivity and citation counts. The results demonstrated that the increase in linguistic positivity is a common phenomenon across the examined academic disciplines. In addition, hard disciplines showed a higher and faster-growing degree of linguistic positivity compared with soft disciplines. Last, a significant positive correlation was identified between citation counts and the degree of linguistic positivity. Reasons for the temporal dynamics and disciplinary variation of linguistic positivity were explored, and implications for the scientific community were discussed.

Can ChatGPT be used to predict citation counts, readership, and social media interaction? An exploration among 2222 scientific abstracts

“…These include basic textual features, such as paper metadata (Ma et al, 2021) and paper length (Haustein et al, 2015;Xie et al, 2019). Additionally, more refined textual analyses have been proposed as citation predictors, encompassing linguistic complexity and readability metrics (Ante, 2022;Lu et al, 2019;Sienkiewicz & Altmann, 2016), stylometry of titles and abstracts (Jimenez et al, 2020), and sentiment analysis (Liu & Zhu, 2023). In a review regarding the application of artificial intelligence for predicting citations and other research assessment metrics, Kousha and Thelwall (2022) observed that the existing predictors, such as readability scores and text-based analyses, have inconsistent predictive power.…”