Computing the semantic similarity between pairs of terms plays a vital role within a myriad of shared data applications, such as data integration and ontology evolution. A first step towards building such applications is to determine which terms are semantically similar to each other. One feasible way to compute the similarity of two terms is to assess their word similarity by exploiting different knowledge resources, e.g., ontologies or domain corpora. Recently, information-theoretic approaches have shown promising results by computing the information content of concepts from the knowledge provided by ontologies. In these methods, the Most Informative Common Subsumers MICSs of two concepts play an important role in determining their similarity. While intuitive, surprisingly, these approaches often do not mirror human judgement well. In this paper, we investigate the effect of choosing a suitable subsumer, called Consensus Common Subsumer (CCS) among all common ancestors of two concepts, on the quality of the term similarity assessment. We cast this issue as an optimization problem by adopting the Particle Swarm Optimization algorithm as one of the most capable optimization approaches. An empirical evaluation based on well-established biomedical benchmarks and ontologies illustrates the accuracy of the proposed approach compared to state-of-the-art approaches.