Tackling the vocabulary mismatch has been a long-standing and major goal in information retrieval. e state-of-the-art solutions mainly rely on leveraging either the relational semantics provided by external resources or the distributional semantics, recently investigated by deep neural approaches. Guided by the intuition that the relational semantics might improve the eeectiveness of deep neural approaches, we propose the Deep Semantic Resource Inference Model (DSRIM) that relies on a twofold contribution: 1) a representation of raw-data that models the relational semantics within text representations by jointly considering objects and relations expressed in a knowledge resource, and 2) an end-to-end neural architecture that jointly learns the query-document relevance and the combined distributional and relational semantic representation of documents and queries. e experimental evaluation is carried out on two TREC datasets from TREC Terabyte and TREC CDS tracks, and two diierent knowledge resources, respectively Word-Net and MeSH. e results indicate that knowledge resource-driven representations allow obtaining similar representations for similar documents while discriminating non-similar documents. Also, we show that our model outperforms state-of-the-art semantic and deep neural information retrieval models.