Identifying the semantic relations that hold between words is of crucial importance for reasoning purposes. Within this context, different methodolo-gies have been proposed that either exclusively focus on a single lexical relation (two-class problem) or learn specific classifiers capable of identifying multiple semantic relations (multi-class problem). In this paper, we propose another way to look at the problem that relies on the multi-task learning paradigm. Preliminary results based on simple learning strategies and state-of-the-art distributional feature representations show that concurrent learning can lead to improvements.