As a balanced indicator for both the quality and productivity of researchers’ scientific outputs, h-index suffers severely from loss of significant information in citation distributions so that its ability to discriminate researchers is rather limited. H-index not only fails to distinguish researchers with quasi the same h-values, but also lacks the ability to promote researchers’ foci on originality and high quality. Current extensions to h-index by exploiting citation distribution characteristics all exhibit some “pathologically” eccentric behaviors against either real datasets or theoretical distributions. With a motivation of balancing quality and quantity, this paper proposes a new h-type index h*pt by exploiting citation distribution characteristics, which encourages quality by defining reward/penalty factors to excess/long-tail citations, and at the same time balances productivity by scaling citations outside the h-core to the average performance of the long-tail papers. Extensive experiments on both the ACL Anthology Network dataset and several theoretical citation distributions prove that the proposed h*pt achieves reasonably better results than previous h-type extensions.
|Translated title of the contribution||h*pt: An h-type index encouraging quality and balancing productivity|
|Original language||Chinese (Simplified)|
|Number of pages||13|
|Publication status||Published - Jan 2017|
- citation distribution
- research evaluation
- qualitative indicator