Analyzing the relationship between text features and grants productivity

On predicting research grants productivity

“…While the results are significant, the typical prediction accuracy were not very high. They were typically higher, though, than approaches based on textual features alone (Tohalino et al 2021). Our analysis also revealed that both Support Vector Machines and Multilayer Perceptron were the classifiers yielding the highest accuracy rates.…”

“…They found that larger abstracts comprising a low number of common words are among the main patterns associated with larger funding values. In a similar study, Tohalino et al (2021) found that topical and complexity textual features play a role in grants predicting grants productivity, but the prediction values were not very high.…”

“…More recently, several studies have focused on analyzing the factors underlying grants success (Boyack et al 2018, Letchford et al 2016, Paiva et al 2012, Tohalino et al 2021.…”

“…Text-and reference-based features have been used to predict the success of research proposals (Boyack et al 2018, Letchford et al 2016, Paiva et al 2012, Tohalino et al 2021. Boyack et al (2018) found that proposals success depends on the topic being approached.…”

“…We used several features such as total number of publications, citations, relevance of PI's institution and diversity of the approached subjects. Given the unbalanced nature of the dataset, we considered as criteria for productivity the publication of at least one scientific paper (Tohalino et al 2021). Using a dataset of research grants from the São Paulo Research Foundation (FAPESP-Brazil), our analysis was conducted in three distinct research areas, namely Medicine, Dentistry and Veterinary Medicine.…”