Relation Classification via Multi-Level Attention CNNs

• DOI: 10.18653/v1/p16-1123
• Publication Date: 2016-08-13T19:45:33Z
• Authors: Linlin Wang; Zhu Cao; Gerard de Melo; Zhiyuan Liu
• Keywords: 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology

Relation classification is a crucial ingredient in numerous information extraction systems seeking to mine structured facts from text.We propose novel convolutional neural network architecture for this task, relying on two levels of attention order better discern patterns heterogeneous contexts.This enables endto-end learning task-specific labeled data, forgoing the need external knowledge such as explicit dependency structures.Experiments show that our model outperforms previous state-of-the-art methods, including those much richer forms prior knowledge.