Author Matching Across Different Academic Databases: Aggregating Simple Feature-Based Rankings

摘要

Profiling the research accomplishments (e.g., papers, research grants, awards, and dissertations) of individual researchers is needed for evaluating their research activity and academic networking. These research accomplishments are generally scattered across separate databases that have different schema. This paper explores an approach for author matching across different academic databases, automatically integrating the records and regarding them as an individual researcher's accomplishments. Given an author identifier with a certain full name in a source database and its counterpart candidates with the same full name in a different target database, we first extract six types of simple features based on the attributes that are easily available in any type of scholarly contribution. Each feature ranks all the namesakes according to similarity with the target author. Finally, we apply unsupervised aggregation to all of the ranked lists, providing an improved ranked list to help manual inspections. In experiments that match researchers in Japan to their PhD dissertations, we demonstrate that the proposed aggregated ranking achieved the best performance over single rankings.