基于多标签策略的中文知识图谱问答系统研究

现有多数中文知识图谱问答（CKBQA）系统侧重于回答单个三元组查询的简单问题，而不能有效解决涉及多个实体和关系的复杂问题。提出一种基于多标签策略进行答案搜索的CKBQA系统，该系统主要包括问题处理和答案搜索2个部分。在问题处理部分，结合预训练语言模型构建新的模型框架，对问题进行实体提及识别、实体链接和关系抽取处理，通过设置3种分类标签将问题划分为简单问题、链式问题和多实体问题。在答案搜索部分，对上述3种分类问题分别给出不同的解决方法。实验结果表明，该系统在CCKS2019-CKBQA评测数据验证集上的平均F1值可达66.76%。

Study of Chinese Knowledge Base Question Answering System Based on Multi-Label Strategy

Most of the existing Chinese Knowledge Base Question Answering(CKBQA) system focus on simple questions that need a single triplet query,but cannot solve complex questions involving multiple entities and relations.To address the problem,this paper proposes a CKBQA system for answer search based on multi-label strategy. The system mainly consists of two parts:question processing and answer search.In the question processing part,a new model framework is constructed based on the pre-trained language model to perform entity mention recognition,entity linking and relation extraction for the questions. By setting three classification labels,the questions are divided into simple questions,chain questions and multi-entity questions.In the answer search part,different processing methods are given for the above three kinds of classification questions. The experimental results show that the average F1 value of the proposed system reaches66.76% on the validation set of evaluation data,CCKS2019-CKBQA.