Quantifying and recommending expertise when new skills emerge

Abstract

In the rapidly changing technological world of today, new technical areas emerge quickly, and new skills related to them garner high demand. In this paper, our goal is to recommend experts for new skills and skill topics. We propose multiple predictive models to utilize data from different enterprise sources: employee assessment data, free-text skill description data, and employee tags from corporate social media. These models include collaborative filtering, content-based, and novel hybrid recommendation approaches. We apply them in an empirical study of real-world corporate data, in which we compare and contrast the models to gain insight on the drivers of performance. The considered data is both structured and unstructured, messy, subjective, and incomplete. The central theme of the paper is to understand how to use data from different sources and what each data source contributes in the expertise management domain. © 2013 IEEE.