Signed directed networks with positive or negative links convey rich information such as like or dislike, trust or distrust. Existing work of sign prediction mainly focuses on triangles (triadic nodes) motivated by balance theory to predict positive and negative links. However, real-world signed directed networks can contain a good number of bridge edges which, by definition, are not included in any triangles. Such edges are ignored in previous work, but may play an important role in signed directed network analysis. In this paper, we investigate the problem of learning representations for signed directed networks. We present a novel deep learning approach to incorporating two social-psychologic theories, balance and status theories, to model both triangles and bridge edges in a complementary manner. The proposed framework learns effective embeddings for nodes and edges which can be applied to diverse tasks such as sign prediction and node ranking. Experimental results on three real-world datasets of signed directed social networks verify the essential role of "bridge" edges in signed directed network analysis by achieving the state-of-the-art performance.