Topic modeling is a technique used to identify and organize topic domains in document collections. Therefore, topic models are extensively used in expert finding by considering domain (expertise) as a topic or a list of topics. Most of the existing topic models, for expert/expertise finding, make use of different versions of Latent Dirichlet Allocation (LDA) to acquire coherent domains. However, in the process of obtaining domains with higher coherence, LDA fails in discovering latent relationships among those domains. In this work, we present our model that combines LDA with Sentence-BERT (S-BERT), further refining the domains with clustering. More specifically, we propose fuzzy clustering using Fuzzy C-means in order to discover latent relationships among the domains. Furthermore, we add temporal information into our model to analyze the relationship between domain word semantics and temporal factor. We compare our model with baselines using the COVID-19 Open Research Dataset (CORD-19) and 20Newsgroups dataset. Through experiments, we demonstrate that our fuzzy topic model outperforms baseline models on both datasets.