摘要
知识库问答(KBQA)是利用知识库中的实体和关系等事实信息来回答自然语言问题的一种方法。在知识库问答中,复杂问题往往存在着多跳和约束。以前的方法是分别对查询图和问题进行编码,并根据谓词序列的文本相似性对它们进行比较。然而,这些方法都忽略了问题和查询图不同的结构,缺乏对不同的结构的处理以及对问题语义结构的考虑。论文提出了一个基于语义图的问题编码模型。采用骨架解析的依赖解析方法将问题转换成图结构,并使用基于图注意力机制的关系上下文的模型对其进行编码。同时捕获问题的文本信息和语义结构信息。最后通过实验分析,验证了论文的方案是可行的和有效的。
Knowledge base ouestion and answer(KBQA)is an approach to answering natural language questions using factual information such as entities and relationships in a knowledge base.In knowledge base question and answer,complex questions often have multiple hops and constraints.Previous approaches have approached this by encoding the query graph and the question separately and comparing them based on the textual similarity of the predicate sequence.However,these approaches have ignored the different structures of questions and query graphs,lacking a treatment of the different structures as well as lacking consideration of the semantic structure of the questions.This paper proposes a model for encoding questions based on semantic graphs.A skeleton parsing dependency resolution approach is used to transform the problem into a graph structure and encode it using a model of relational context based on the graph attention mechanism.Both the textual and semantic structure information of the problem is captured.Finally,the experimental analysis verifies that the scheme of this paper is feasible and effective.
作者
万恒
田萍芳
WAN Heng;TIAN Pingfang(School of Computer Science and Technology,Wuhan University of Science and Technology,Wuhan 430065;Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System,Wuhan University of Science and Technology,Wuhan 430065)
出处
《计算机与数字工程》
2025年第6期1675-1680,共6页
Computer & Digital Engineering
关键词
查询图
知识图谱
知识库问答
复杂问题
图注意力网络
query graph
knowledge graph
knowledge base question answering
complex question
graph attention networks
