Scholarly profiles, user preferences and impact scores

Abstract

See video of the presentation.

We investigate the digital presence of scholars at different academic Web sites. With new technologies, creating profiles, disseminating and exchanging ideas is easily done, and scholars are more likely to attend the networks and impact their community.

In our study we compare research profiles of employees at the University of Bergen at five different academic network sites. The sites are ResearchGate, Academia.edu, Google Scholar, ResearcherID and ORCID. CRIStin, the Current Research Information System in Norway (www.cristin.no), is used as a reference value. CRIStin is a national database which contains quality-assured data on scientific publications including supplementary author details such as age, gender, position and affiliation. All investigated sites have varying scopes (and degree of control), but also common features which are worth to investigate and compare.

Data is collected using Web scraping applications developed at the University of Bergen Library by searching for the researchers that are affiliated with the University of Bergen.  This was achieved by analyzing the Document Object Model (DOM) of every academic site and then building up a set of selectors and expressions, so that the DOM could be traversed programmatically and indicators extracted.

Author recognition is then done by comparing names given in the services with names in CRIStin. After extensive data cleansing and deduplication we were able to compare the different services.

Our first goal is to determine number of profiles and degree of overlap. The overlap tells us whether scholars are willing to maintain their profiles at several services. Preference of platform in regard to faculty affiliation, position and age is another aspect of our investigation.

Further, we analyze extracted indicators in regard to traditional bibliometric and “altmetric” measures. Bibliometric measures are related to publications and citations, while “altmetric” indicators comprise different forms of Web activities such as followers, following, views and downloads. The indicators vary from service to service, and a correlation analysis tells us whether indicators are related to each other or not.

We find that about 37% of researchers at the University of Bergen have at least one profile. They are reluctant to maintain several profiles and overlap was therefore relatively small. Age is a poor predictor of web site use, and women are underrepresented on the investigated platforms. The representation is highest at the Faculty of Psychology and the Faculty of Social Sciences (> 40%). Available indicators show high correlation within bibliometric indicators, but correlation is weak with social and activity indicators across platforms.