Deep ranking based cost-sensitive multi-label learning for distant supervision relation extraction

Highlights：

• In distant supervision relation extraction, the relation classes can have latent connections (class ties). In this paper, to exploit class ties between relations, we propose a general ranking based multi-label learning framework combined with convolutional neural networks. Ranking based loss functions with regularization technique are introduced to learn the latent connections between relations. Furthermore, to relieve the class-imbalance problem in distant supervision relation extraction, we further adopt cost-sensitive learning to rescale the costs from the positive and negative labels. Extensive experiments on a widely used dataset show the effectiveness of our model to exploit class ties and to relieve class imbalance problem. Our model achieves state-of-the-art performance.