Modeling of cross-disciplinary collaboration for potential field discovery and recommendation based on scholarly big data

Abstract

The promise of cross-disciplinary scientific collaboration has recently been proven by both technological innovation and scientific research. Much effort has been spent on research collaboration recommendation. A remaining challenge is to make valuable recommendation to specific researchers in specific fields in order to obtain more fruitful cross-disciplinary collaboration. Cross-disciplinary information hides in big data and the relationships between different fields are complicated, complex, and subtle. This paper proposes a method for cross-disciplinary collaboration recommendation (CDCR) to analyze cross-disciplinary collaboration patterns in scholarly big data, and recommend valuable research fields for possible cross-disciplinary collaboration. A cross-disciplinary discovery algorithm based on topic modeling is designed to extract potential research fields. Collaboration patterns are examined by analyzing the research field correlations. A recommendation algorithm is developed to provide a specific recommendation list of potential research fields according to the discovered cross-disciplinary collaboration patterns with researchers’ profiles. Evaluations conducted based on a real scholarly dataset demonstrate the effectiveness of the proposed method in recommending potentially valuable collaborations.