| 299 | 2 | 75 |
| 下载次数 | 被引频次 | 阅读次数 |
组织和检索信息是人机交互重点关注的话题之一。基于知识图谱(Knowledge Graph, KG)的智能问答系统通过语义解析用户问题，检索知识并回答问题，已成为一种信息检索的有效途径，是人机交互的典型应用。时序知识图谱(Temporal Knowledge Graph, TKG)问答系统通过语言模型获取问题中的实体和时间戳，并在大型TKG中检索答案。TKG问答系统包含2个挑战：(1)给定问题，需检索整个TKG,效率低且易受干扰项的影响；(2)难以捕获问题中隐含的时间词和时间顺序信息。提出一种基于图注意力网络的时间对比学习(Time Contrast Learning, TCL)模型，将源问题与替换时间词后的对比问题同时训练，使用图注意力网络更新实体邻接子图的节点特征，缩小潜在答案的检索空间。在CRONQUESTIONS数据集上进行大量实验，结果表明TCL比其他基准方法具有更好的性能，相较于最先进的基准方法在Hit@1和Hits@10指标上平均提升3.44%和2.02%。
Abstract：Organizing and retrieving information is one of the key topics in human-computer interaction. Intelligent question-answering system over Knowledge Graph(KG) has become an effective way to retrieve information through semantic parsing of user's questions, retrieving knowledge and answering questions, which is a typical application of human-computer interaction. Question-answering system over Temporal Knowledge Graph(TKG) obtains entities and timestamps from questions through the language model and retrieves answers from a large TKG However, the question-answering system over temporal knowledge graphs contains two challenges:(1) Given the question, the entire TKG needs to be retrieved, which is inefficient and vulnerable to interfering items;(2) It is difficult to capture the time words and chronological information in the question. A Time Contrast Learning(TCL) is proposed, which trains both the source problem and the contrast problem after replacing the temporal words, and uses the graph attention network to update the node features of the entity neighbor sub-graph, in order to reduce the retrieval space for potential answers. Extensive experiments are conducted on the CRONQUESTIONS dataset, and the results show that TCL has better performance than other benchmark methods with an average improvement of 3.44% in the metric of Hits@1 and 2.02% in Hits@10,respectively, when compared to the state-of-the-art baseline method.
[1] 姚元杰，龚毅光，刘佳，等.基于深度学习的智能问答系统综述[J].计算机系统应用，2023,32(4):1-15.
[2] BORDES A,USUNIER N,GARCIA-DURAN A,et al.Translating Embeddings for Modeling Multi-relational Data[C]//Proceedings of the 27th Annual Conference on Neural Information Processing Systems.Lake Tahoe:Curran Associates Inc.,2013:2787-2795.
[3] TROUILLON T,WELBL J,RIEDEL S,et al.Complex Embeddings for Simple Link Prediction[C]//Proceedings of the 33rd International Conference on Machine Learning.New York:JMLR,2016:2071-2080.
[4] SUN Z Q,DENG Z H,NIE J Y,et al.RotatE:Knowledge Graph Embedding by Relational Rotation in Complex Space[C]//Proceedings of the 7th International Conference on Learning Representations.New Orleans:[s.n.],2019:1981-1999.
[5] JIANG T S,LIU T Y,GE T,et al.Towards Time-aware Knowledge Graph Completion[C]//Proceedings of COLING 2016,the 26th International Conference on Computational Linguistics:Technical Papers.Osaka:The COLING 2016 Organizing Committee,2016:1715-1724.
[6] DASGUPTA S S,RAY S N,TALUKDAR P.HyTE:Hyperplane-based Temporally Aware Knowledge Graph Embedding[C]//Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing.Brussels:ACL,2018:2001-2011.
[7] GARCíA-DURáN A,DUMAN■.Learning Sequence Encoders for Temporal Knowledge Graph Completion[C]//Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing.Brussels:ACL,2018:4816-4821.
[8] LACROIX T,OBOZINSKI G,USUNIER N.Tensor Decompositions for Temporal Knowledge Base Completion[C]//Proceedings of the 8th International Conference on Learning Representations.Addis:[s.n.],2020:3057-3069.
[9] GOEL R,KAZEMI S M,BRUBAKER M,et al.Diachronic Embedding for Temporal Knowledge Graph Completion[C]//Proceedings of the AAAI Conference on Artificial Intelligence.New York:AAAI,2020:3988-3995.
[10] JIN W,KHANNA R,KIM S,et al.ForecastQA:A Question Answering Challenge for Event Forecasting with Temporal Text Data[C]//Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing.Online:ACL,2021:4636-4650.
[11] NING Q,WU H,HAN R J,et al.TORQUE:A Reading Comprehension Dataset of Temporal Ordering Questions[C]//Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing.Online:ACL,2020:1158-1172.
[12] 乔少杰，杨国平，于泳，等.QA-KGNet:一种语言模型驱动的知识图谱问答模型[J].软件学报，2023,34(10):4584-4600.
[13] 萨日娜，李艳玲，林民.知识图谱推理问答研究综述[J].计算机科学与探索，2022,16(8):1727-1741.
[14] JIA Z,ABUJABAL A,ROY R S,et al.TempQuestions:A Benchmark for Temporal Question Answering[C]//Proceedings of Companion of the Web Conference 2018.Lyon:ACM,2018:1057-1062.
[15] JIA Z,ABUJABAL A,ROYS R,et al.TEQUILA:Temporal Question Answering over Knowledge Bases[C]//Proceedings of the 27th ACM International Conference on Information and Knowledge Management.Torino:ACM,2018:1807-1810.
[16] JIA Z,PRAMANIK S,ROY R S,et al.Complex Temporal Question Answering on Knowledge Graphs[C]//Proceedings of the 30th ACM International Conference on Information and Knowledge Management.Queensland:ACM,2021:792-802.
[17] SAXENA A,CHAKRABARTI S,TALUKDAR P.Question Answering over Temporal Knowledge Graphs[C]//Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing.Online:ACL,2021:6663-6676.
[18] SAXENA A,TRIPATHI A,TALUKDAR P.Improving Multi-hop Question Answering over Knowledge Graphs Using Knowledge Base Embeddings[C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics.Online:ACL,2020:4498-4507.
[19] LIU Y H,LIANG D,FANG F,et al.Time-aware Multiway Adaptive Fusion Network for Temporal Knowledge Graph Question Answering[EB/OL].(2023-02-24)[2023-04-17].https://arxiv.org/abs/2302.12529.
[20] BRODY S,ALON U,YAHAV E.How Attentive are Graph Attention Networks?[EB/OL].(2021-05-30)[2023-05-10].https://arxiv.org/abs/2105.14491.
[21] 俞海亮，彭冬亮，谷雨.结合双层多头自注意力和BiLSTM-CRF的军事武器实体识别[J].无线电工程，2022,52(5):775-782.
[22] 梁礼明，何安军，阳渊，等.融合Transfomer和多尺度并行注意的结直肠息肉分割算法[J].无线电工程，2023,53(1):209-219.
[23] FéVRY T,SOARES L B,FITZGERALD N,et al.Entities as Experts:Sparse Memory Access with Entity Supervision[EB/OL].(2020-04-15)[2023-05-10].https://arxiv.org/abs/2004.07202.
基本信息:
DOI：
中图分类号:TP18;TP391.1
引用信息:
[1]于泳,乔少杰,陈金勇等.基于图注意力网络的时序知识图谱人机交互模型[J].无线电工程,2024,54(07):1676-1686.
基金信息:
国家自然科学基金(62272066); 四川省科技计划(2021JDJQ0021,2022YFG0186,2022NSFSC0511,2023YFG0027); 教育部人文社会科学研究规划基金(22YJAZH088); 宜宾市引进高层次人才项目(2022YG02); 成都市“揭榜挂帅”科技项目(2022-JB00-00002-GX,2021-JB00-00025-GX); 成都市技术创新研发项目(重点项目)(2024-YF08-00029-GX); 成都市区域科技创新合作项目(2023-YF11-00020-HZ); 中国电子科技集团公司第五十四研究所高校合作课题(SKX212010057); 成都海关科研项目资助(2022CK008)~~