We investigate a new approach to search results diversification aimed at the query that we call "Diversified Query Expansion" (DQE). We propose a new term-level DQE method in which word similarity is determined at the term level based on external resources. Then, we investigate a new method for query-level resource weighting based on a set of features to generate expansion terms proportional to the weight of that resource. Finally, we introduce a novel method for aspect-level DQE which relies on an explicit modeling of query aspects based on embedding in order to select expansion terms at a latent (semantic) level to cover more aspects of a given query.