Tracking expertise profiles in community-driven and evolving knowledge curation platforms

Abstract

Acquiring and managing expertise profiles represents a major challenge in any organization, as often, the successful completion of a task depends on finding the most appropriate individual to perform it. User profiling has been extensively utilised as a basis for recommendation, personalisation and matchmaking systems. Accurate user profile generators can improve interaction and collaboration between researchers working in similar domains but in different locations or organizations. They can also assist with identifying the optimum set of researchers with complementary skills for cross-disciplinary research teams at a given time. The topic of expertise modelling has been the subject of extensive research in two main disciplines: Information Retrieval (IR) and Social Network Analysis (SNA). Traditional IR and SNA expertise profiling techniques rely on large corpora of static documents authored by an expert, such as publications, reports or grants, the content of which remains unchanged due to the static and final nature of such resources. Consequently, such techniques build the expertise model through a document-centric approach that provides only a macro-perspective of the knowledge emerging from such documents. With the emergence of Web 2.0, there has been a significant increase in online collaboration, giving rise to vast amounts of accessible and searchable knowledge in platforms where content evolves through individuals’ contributions. This increase in participation provides vast sources of information, from which knowledge and intelligence can be derived for modelling the expertise of contributors. However, with the proliferation of collaboration platforms, there has been a significant shift from static to evolving documents. Wikis or collaborative knowledge bases, predominantly in the biomedical domain, support this shift by enabling authors to incrementally and collaboratively refine the content of the embedded documents to reflect the latest advances in knowledge in the field. Regardless of the domain, the content of these living documents changes via micro-contributions made by individuals, thus making the macro-perspective, provided by the document as a whole, no longer adequate for capturing the evolution of knowledge or expertise. Hence, expertise profiling is presented with major challenges in the context of dynamic and evolving knowledge. Thus, the shift from static documents to living documents requires a shift in the way in which expertise profiling is performed. This thesis examines methods for advancing the state of the art in expertise modelling by considering dynamic content; i.e., platforms in which, knowledge evolves through micro-contributions. Towards this goal, a novel expertise profiling framework is introduced that provides solutions for expertise modelling in the context of platforms where knowledge is subject to continuous evolution through experts’ micro-contributions; i.e., given a series of micro-contributions, the aim is to build an expertise profile for the author of those micro-contributions. Furthermore, as the expertise of an individual is dynamic and usually changes with time, the proposed framework aims at capturing the temporality of expertise, in order to facilitate tracking and analysis of changes in interests and expertise over time. The proposed framework comprises three major elements: (i) a model, aimed at capturing the fine-grained provenance of micro-contributions and evolving content in the macro-context of the host living documents, as well as the temporality of micro-contributions; (ii) a domain-independent methodology for building expertise profiles by capturing expertise topics in micro-contributions and consolidating them to weighted concepts from domain ontologies, and (iii) a profile refinement mechanism for complementing expertise profiles by integrating contextual factors in existing social expert networks. Furthermore, the proposed expertise profiling framework creates profiles containing ontological concepts, each of which represents an area of expertise. This provides the flexibility of using the structure of domain ontologies to represent the expertise topics embedded in the micro-contributions of an expert, at different levels of granularity. In addition, using ontological concepts to represent expertise topics facilitates the use of semantic similarity for comparing profiles that describe expertise at different levels of abstraction. This in turn facilitates the semantic evaluation of expertise profiles, rather than evaluation based on the exact matching of concepts or terms. Moreover, using the structure of ontologies allows experts to customise the granularity of their profiles in order to complement their existing profiles with fine-grained domain concepts representing knowledge embedded in their micro-contributions to evolving knowledge-curation platforms. Finally, this thesis presents the Profile Explorer visualization tool, which serves as a paradigm for exploring and analysing time-aware expertise profiles in knowledge bases where content evolves over time. Profile Explorer facilitates browsing, search and comparative analysis of evolving expertise, independent of the domain and the methodology used in creating profiles.