We present a novel method for relation extraction (RE) from a single\nsentence, mapping the sentence and two given entities to a canonical fact in a\nknowledge graph (KG). Especially in this presumed sentential RE setting, the\ncontext of a single sentence is often sparse. This paper introduces the KGPool\nmethod to address this sparsity, dynamically expanding the context with\nadditional facts from the KG. It learns the representation of these facts\n(entity alias, entity descriptions, etc.) using neural methods, supplementing\nthe sentential context. Unlike existing methods that statically use all\nexpanded facts, KGPool conditions this expansion on the sentence. We study the\nefficacy of KGPool by evaluating it with different neural models and KGs\n(Wikidata and NYT Freebase). Our experimental evaluation on standard datasets\nshows that by feeding the KGPool representation into a Graph Neural Network,\nthe overall method is significantly more accurate than state-of-the-art\nmethods.\n