Expertise Identification in Enterprise Social Media

Abstract

The increasing adoption of Enterprise Social Media (ESM) systems within enterprises is driven by the need for the explicit facilitation of sharing expertise. Expertise Identification (EI) functionality can satisfy this need. The social-media-like content and Collaborative Filtering (CF) annotation data available in ESM, however, pose unique requirements on EI. In this light, we perform an elaborate study into literature and practice surrounding ESM, expertise, and EI, in order to formulate a number of design requirements and choices for EI in ESM. In our case study on E-view, a live ESM system, we design, implement, and test an EI prototype that stores all ESM relationships in a social graph and all user content into a search engine, which are then combined to estimate user expertise scores. Our results reveal that relevant content used to estimate expertise scores should be selected on the basis of both full-content and tags. Due to the sparsity of CF appreciation data in the dataset, EI strategies that complement content relevance scoring with appreciation scoring for the estimation of expertise scores, perform equally well as strategies based only on content relevance, in terms of ranked lists of experts. As such, we recommend that future work retests the EI strategies with the constructed prototype, using an ESM dataset that contains more CF appreciation data. We present a number of recommendations for EI in ESM and for reusing our evaluation methods in future research.