Discourse relation recognition method and device based on knowledge distillation and multi-task learning

The invention provides a discourse relation recognition method and device based on knowledge distillation and multi-task learning, and the method comprises the steps: on one hand, sharing knowledge between a conjunction classification auxiliary task and an implicit discourse relation recognition main task based on a parameter sharing mode; on the other hand, knowledge in the teacher model with theenhanced conjunctions is migrated to the corresponding implicit discourse relation recognition model from the feature layer and the classification layer based on the knowledge distillation technology; and therefore, the recognition performance of the student model is improved by fully utilizing the conjunction information inserted during corpus annotation. 本发明提出一种基于知识蒸馏和多任务学习的篇章关系识别方法与装置，在本发明中，一方面，基于参数共享的方式在连接词分类辅助任务和隐式篇章关系识别主任务之间共享知识；另一方面，将连接词增强的教师模型中的知识基于知识蒸馏技术从特征层和分类层迁移到相应的隐式篇章关系识别模型中；以充分利用语料标注时插入的连接词信息提高学生模型的识别性能。