Detecting User Profiles in Collaborative Ontology Engineering Using a User’s Interactions

Abstract

Collaborative ontology-engineering methods usually prescribe a set of processes, activities, stakeholders, and the roles each stakeholder plays in these activities. We, however, believe that (a) the stakeholder community of each ontology-engineering project is different, and (b) one can observe different types of user behavior. It may thus very well be that the prescribed set of stakeholder types and roles do not suffice. If one were able to identify these user behavior types, which we will call user profiles, one can compliment or revisit those predefined roles. For instance, those user profiles can be used to provide customized interfaces for optimizing activities in certain ontology-engineering projects. We present a method for discovering different user profiles based on the interactions users have with each other on a collaborative ontology-engineering environment. Our approach clusters users based on the types of interactions they perform, which are retrieved from datasets that were annotated with an interaction ontology—built on top of SIOC—that we have developed. We demonstrate our method using the database of two instances of the GOSPL ontology-engineering tool. The databases contain the interactions of two distinct ontology-engineering projects involving, respectively, 42 and 36 users. For each dataset, we discuss the findings by analyzing the different clusters. We found that we are able to discover different user profiles, indicating that the approach we have taken is viable, though more experiments are needed to validate the results and to discover patterns across ontology-engineering projects.