Leveraging Institutional Data for Author Name Disambiguation

Author name disambiguation is a challenging problem in computer science. The problem arises from the fact that many authors share similar or identical names. Although some scholarly databases assign unique author identifiers, levels of accuracy are often unacceptable—especially for authors with common names. Existing algorithms have largely not leveraged institutional data on individual researchers. We are extending ReCiter, an agglomerative clustering algorithm for author name disambiguation, for use in publication management at our institution. The system uses available institutional data on researchers, including primary and secondary departments, history of co-investigatorships on grants and co-authorships, favored journals, and years of authors' terminal academic degrees. We are investigating the use of machine learning approaches to optimize system performance, and are planning to make the system available as a suite of freely available, open-source tools.