Searching for associations between entities is needed in many areas. On the Semantic Web, it usually boils down to finding paths that connect two entities in an entity-relation graph. Given the increasing volume of data, apart from the efficiency of path finding, recent research interests have focused on how to help users explore a large set of associations that have been found. To achieve this, we propose an approach to exploratory association search, called Explass, which provides a flat list (top-K) of clusters and facet values for refocusing and refining the search. Each cluster is labeled with an ontological pattern, which gives a conceptual summary of the associations in the cluster. Facet values comprise classes of entities and relations appearing in associations. To recommend frequent, informative, and small-overlapping patterns and facet values, we exploit ontological semantics, query context, and information theory. We compare Explass with two existing approaches by conducting a user study over DBpedia, and test the statistical significance of the results.