Neural Classification of Research Scientific Excellence in Universities

Universities strive to conduct research of the highest scientific quality and social impact to drive innovation, push the boundaries of knowledge, achieve satisfactory contribution to society, meet international standards, strengthen their reputation, and help win the fight for competitive funding of activities. Sponsors usually prioritize research project concepts that include declarations of achieving scientific excellence and the societal impact of research. In this respect, multicriteria assessments of frontier or high-risk research pose a significant challenge, and expert peer reviews play a key role. Especially when it is necessary to consider quantitative and qualitative indicators, e.g., regarding differently understood social benefits, it is desirable to provide proposals of methods and tools that support decision-making processes in the complex classification of different research concepts within the usually intense competition for very limited funds. This work employs selected neural models to classify research concepts in terms of excellence assessed by multiple criteria. The results of theoretical analyses and experiments based on empirical data prove that such models can simplify evaluation processes and provide significant support for experts, who can make it easier to make objective decisions and avoid accusations of bias. The advantages and disadvantages of individual models and challenges related to future directions of their improvement are presented.