[EMNLP2017]Global Normalization of Convolutional Neural Networks for Joint Entity and Relation

每次输入的一对实体，输出就是两个实体的类型以及两个实体的关系．不能称之为全局优化的的方法，全局的优化可以借鉴＜End-to-End Neural Relation Extraction with Global Optimization＞