【摘要】 基于知识图谱的问答系统是一种新兴的问答模式,能够用准确的答案回答用户提出的事实类问题,但目前广泛使用的基于深度学习技术的问答系统无法在缺少标注数据的情景下工作。针对这一问题,本文提出了基于问题生成的知识图谱问答方法。该方法无需成本高昂的标注数据,编写简单的模版文件即可构建问答系统。在自建哈尔滨工业大学学校信息数据集上测试,相比于深度学习的基准方法,本方法在无标注数据情景下具有可用性。更多还原

【Abstract】 The knowledge graph based question answering system is a new question answering mode,which can answer the fact questions raised by users with accurate answers. The widely used question answering system based on deep learning technology cannot work in the absence of labeled data. Aiming at this problem,this paper proposes a question answering method based on question generation. This method does not require costly annotation data,and a simple template file can be written to construct a question answering system. On the self-built Harbin Institute of Technology school information data set,compared with the deep learning benchmark method,the test results show the usability of this method without labeling data.更多还原