Assessing Professional Skills in a Multi-scale Environment by Means of Graph-Based Algorithms

Abstract

The present paper introduces a study of different techniques to assess professional skills in social networks and to align those user skills with existing multi-scale knowledge classifications. Currently both job seekers and talent hunters are looking for new and innovative techniques to filter jobs and candidates as well as candidates are also trying to improve and make more attractive their profiles. In this environment it is necessary to provide new techniques to assess the quality of professional skills depending on user's activity and to compare with existing scales. To do so some relevant graph-based techniques such as the HITS and the SPEAR algorithms have been used for calculating the confidence of a certain user in a particular skill. Moreover a new re-interpretation of the SPEAR algorithm, called Skill rank, is introduced to take advantage of user's behavior and history. A major outcome of this approach is that expertise and experts can be detected, verified and ranked using a suited trust metric. The paper also presents a validation of the Skill rank accuracy by means of a sound qualitative and quantitative comparison with existing approaches based on the opinions of a panel of experts (3) on a real dataset (created using the Linked in API) and two different scales. Although results show in general low values of accuracy (close to 50% of correct classified skills), the Skill rank technique is more accurate than other techniques to align a user skill in a certain scale of knowledge. Finally some discussion, conclusions and future work are also outlined.