| 引用本文: | 乔少杰,杨国平,于泳,韩楠,覃晓,屈露露,冉黎琼,李贺.QA-KGNet: 一种语言模型驱动的知识图谱问答模型.软件学报,2023,34(10):4584-4600 |
| |
|
| |
|
|
| 本文已被：浏览 3534次 下载 6217次 |
码上扫一扫！ |
|
|
| QA-KGNet: 一种语言模型驱动的知识图谱问答模型 |
|
乔少杰1, 杨国平1, 于泳1,2, 韩楠3, 覃晓4, 屈露露1, 冉黎琼1, 李贺5
|
|
1.成都信息工程大学 软件工程学院, 四川 成都 610225;2.数字媒体艺术四川省重点实验室 (四川音乐学院), 四川 成都 610021;3.成都信息工程大学 管理学院, 四川 成都 610225;4.广西人机交互与智能决策重点实验室 (南宁师范大学), 广西 南宁 530100;5.西安电子科技大学 计算机科学与技术学院, 陕西 西安 710071
|
|
| 摘要: |
| 基于知识图谱的问答系统可以解析用户问题,已成为一种检索知识、自动回答所询问题的有效途径.知识图谱问答系统通常是利用神经程序归纳模型,将自然语言问题转化为逻辑形式,在知识图谱上执行该逻辑形式能够得到答案.然而,使用预训练语言模型和知识图谱的知识问答系统包含两个挑战:(1)给定问答(question-answering,QA)上下文,需要从大型知识图谱(knowledge graph,KG)中识别相关知识;(2)对QA上下文和KG进行联合推理.基于此,提出一种语言模型驱动的知识图谱问答推理模型QA-KGNet,将QA上下文和KG连接起来形成一个工作图,使用语言模型计算给定QA上下文节点与KG节点的关联度,并使用多头图注意力网络更新节点表示.在CommonsenseQA、OpenBookQA和MedQA-USMLE真实数据集上进行实验来评估QA-KGNet的性能,实验结果表明:QA-KGNet优于现有的基准模型,表现出优越的结构化推理能力. |
| 关键词: 知识图谱 预训练语言模型 QA上下文 多头图注意力网络 联合推理 |
| DOI：10.13328/j.cnki.jos.006882 |
| 分类号: |
| 基金项目:国家自然科学基金(61962006);四川省科技计划(2021JDJQ0021,2022YFG0186);四川音乐学院数字媒体艺术四川省重点实验室资助项目(21DMAKL02);成都市技术创新研发项目(2021-YF05-00491-SN);成都市重大科技创新项目(2021-YF08-00156-GX);成都市“揭榜挂帅”科技项目(2021-JB00-00025-GX);成都市软科学研究项目(2021-RK00-00065-ZF,2021-RK00-00066-ZF);广西重大创新驱动项目(桂科AA22068057);四川省社会科学高水平团队项目(2015Z177) |
|
| QA-KGNet: Language Model-driven Knowledge Graph Question-answering Model |
|
QIAO Shao-Jie1, YANG Guo-Ping1, YU Yong1,2, HAN Nan3, QIN Xiao4, QU Lu-Lu1, RAN Li-Qiong1, LI He5
|
|
1.School of Software Engineering, Chengdu University of Information Technology, Chengdu 610225, China;2.Digital Media Art Key Laboratory of Sichuan Province(Sichuan Conservatory of Music), Chengdu 610021, China;3.School of Management, Chengdu University of Information Technology, Chengdu 610225, China;4.Guangxi Key Lab of Human-machine Interaction and Intelligent Decision(Nanning Normal University), Nanning 530100, China;5.School of Computer Science and Technology, Xidian University, Xi'an 710071, China
|
| Abstract: |
| The question-answering system based on knowledge graphs can analyze user questions, and has become an effective way to retrieve relevant knowledge and automatically answer the given questions. The knowledge graph-based question-answering system usually uses a neural program induction model to convert natural language question into a logical form, and the answer can be obtained by executing the logical form on the knowledge graph. However, the knowledge question-answering system by using pre-trained language models and knowledge graphs involves two challenges: (1) given the QA (question-answering) context, relevant knowledge needs to be identified from a large KG (knowledge graph); (2) it isneeded to perform the joint reasoning on QA context and KG. Based on these challenges, a language model-driven knowledge graph question-answering model is proposed, which connects the QA context and KG to form a joint graph, and uses a language model to calculate the relevance of the given QA context nodes and KG nodes, and a multi-head graph attention network is employed to update the node representation. Extensive experiments on the CommonsenseQA, OpenBookQA and MedQA-USMLE real datasets are conducted to evaluate the performance of QA-KGNet and the experimental results show that QA-KGNet outperforms existing benchmark models and exhibits excellent structured reasoning capability. |
| Key words: knowledge graph (KG) pre-trained language model question-answering (QA) context multi-head graph attention network joint reasoning |
|
|
|
|
