Keyword based profile creation using latent dirichlet allocation, domain dictionary and domain ontology / Nor Adzlan Jamaludin

Expert Finding is a field in information retrieval that focuses on finding an expert based on several criteria. Some of the methods that have been applied for expert finding include statistical, machine learning and ontology-based methods. Profile creation is one of the steps or tasks that are required in expert finding, which is the process of capturing and representing the details of experts and users which later can be used for retrieval. An issue that is faced for profile creation in expert finding is that the profiles being created are focused on the details of the experts but not on the users who are searching for these experts. This research explores a profile creation model that creates domain specific keyword-based profiles of users using Latent Dirichlet Allocation, domain dictionary and domain ontology from bookmarks. The domain of agriculture is selected as the case study for this research. The model is implemented in a form of a prototype and is evaluated by comparing how similar the prototype created profiles with manually built ones. From the results and analysis of the research, it is concluded that the method can successfully create domain specific profiles. The significances and contributions of the research include the application of LDA in user profiling, the proposed model, model prototype and the results and findings of the experiments conducted throughout the research.