Using Probabilistic Topic Models in Enterprise Social Software

Enterprise social software (ESS) systems are open and flexible corporate environments which utilize Web 2.0 technologies to stimulate participation through informal interactions and aggregate these interactions into collective structures. A challenge in these systems is to discover, organize and manage the knowledge model of topics found within the enterprise. In this paper we aim to enhance the search and recommendation functionalities of ESS by extending their folksonomies and taxonomies with the addition of underlying topics through the use of probabilistic topic models. We employ Latent Dirichlet Allocation in order to elicit latent topics and use the latter to assess similarities in resource and tag recommendation as well as for the expansion of query results. As an application of our approach we extend the search and recommendation facilities of the Organik enterprise social system.