Understanding Relationship between Scholars' Breadth of Research and Scientific Impact

Many existing metrics to evaluate scholars consider their scientific impact without considering the importance of breadth of research. In this paper, we define a new metric for breadth of research based on the generalized Stirling metric that considers multiple aspects of breadth of research. We extract research topics in computer science using concept extraction and clustering from the literature in the ACM dataset. We then assign authors a distribution over these research topics, from which we calculate scores of breadth of research for each author. We design five simulation experiments that evaluate the ability of a metric to measure breadth of research and use these experiments to compare our new metric to traditional metrics. The results show how these metrics perform in different experiments, concluding that no metric consistently outperforms the others. We test the relationship between our new metric and scientific impact and find a weak correlation between them. Finally, we find that the variation of the metric over time illustrates a possible publication pattern for scholars. Conference Topic Indicators