知识图谱问答(KBQA)

图一:MindMap

语义解析神经网络(Semantic Parsing Neural Network)

目前基于 Semantic Parsing 的 KBQA 主要有 4 种方法:

语义解析(Semantic Parser)过程转化为查询图生成问题的各类方法; 在领域数据集适用的 Encoder-Decoder 模型化解析方法; 基于 Transition-Based 的状态迁移学习的解析方法; 利用 KV-MemNN 进行解释性更强的深度 KBQA 模型;

传统方法

传统的 Semantic Parser 方法主要依赖于预先定义的规则模板,或者利用监督模型对用户 Query 和语义形式化表示(如 CCG [11]、λ-DCS [12])的关系进行学习,然后对自然语言进行解析。

这种方法需要大量的手工标注数据,可以在某些限定领域和小规模知识库(如 ATIS [13]、GeoQuery)中达到较好的效果,但是当面临 Freebase 或 DBpedia 这类大规模知识图谱的时候,往往效果欠佳。

查询图方法(Query Graph)

编解码器方法(Encoder-Decoder)

Transition-Based 方法

Memory Network 方法

论文:https://arxiv.org/pdf/1410.3916.pdf

信息检索(Information Retrieval)

基于搜索排序(IR)的知识图谱问答的工作流: 首先会确定用户 Query 中实体提及词(Entity Mention); 然后链接到 KG 中的主题实体(Topic Entity),并将 Topic Entity 相关的子图(Subgraph)提取出来作为候选答案集合; 然后分别从 Query 和候选答案中抽取特征; 最后利用排序模型对 Query 和候选答案进行建模并预测

此类方法不需要大量人工定义特征或者模板,将复杂语义解析问题转化为大规模可学习问题。

依据特征表示技术不同,分为基于特征工程的方法和基于表示学习的方法。

基于特征工程的方法

《Information Extraction over Structured Data: Question Answering with Freebase 》是该类方法的基础模型,处理流程:

先对问句进行句法分析,并对其依存句法分析结果提取问题词(qword)、问题焦点词(qfocus)、主题词(qtopic)和中心动词(qverb)特征,将其转化为问句特征图(Question Graph) 然后利用 qtopic 在 KG 内提取 Subgraph,并基于此生成候选答案特征图 最后将问句中的特征与候选特征图中的特征进行组合,将关联度高的特征赋予较高的权重,该权重的学习直接通过分类器学习

但这种方法有两点不足: 需要自行定义并抽取特征,而且问句特征和候选答案特征组合需要进行笛卡尔乘积,特征纬度过大; 此方法难以处理复杂问题;

论文:http://cs.jhu.edu/~xuchen/paper/yao-jacana-freebase-acl2014.pdf

基于表示学习的方法

鉴于基于特征工程的两点不足,基于表示学习的方法进行了改进

其他方法

从 KG/QA 中学习模版,然后将问句拆解成 Logical Forms 或理解其意图

把复杂问句分解成几个简单问句,然后从多个简单答案中找出最终的复杂问句答案

Message Parsing NN + Stepwise Reasoning Network

Neural Symbolic Machine(NSM)

开放数据集

WebQuestions

SimpleQuestions

ComplexQuestions

论文集

[1] Yang Z, Qi P, Zhang S, et al. Hotpotqa: A dataset for diverse, explainable multi-hop question answering[J]. arXiv preprint arXiv:1809.09600, 2018.

[2] Saha A, Pahuja V, Khapra M M, et al. Complex sequential question answering: Towards learning to converse over linked question answer pairs with a knowledge graph[C]//Thirty-Second AAAI Conference on Artificial Intelligence. 2018.

[3] Berant J, Chou A, Frostig R, et al. Semantic parsing on freebase from question-answer pairs[C]//Proceedings of the 2013 conference on empirical methods in natural language processing. 2013: 1533-1544. [4] Yih W, Richardson M, Meek C, et al. The value of semantic parse labeling for knowledge base question answering[C]//Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers). 2016: 201-206. [5] Bao J, Duan N, Yan Z, et al. Constraint-based question answering with knowledge graph[C]//Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers. 2016: 2503-2514. [6] 刘康,面向知识图谱的问答系统. AI前沿讲习班第8期PPT [7] 赵军, 刘康, 何世柱, 陈玉博. 知识图谱[B]. 2018. [8] Trivedi P, Maheshwari G, Dubey M, et al. Lc-quad: A corpus for complex question answering over knowledge graphs[C]//International Semantic Web Conference. Springer, Cham, 2017: 210-218. [9] Dubey M, Banerjee D, Abdelkawi A, et al. Lc-quad 2.0: A large dataset for complex question answering over wikidata and dbpedia[C]//International Semantic Web Conference. Springer, Cham, 2019: 69-78. [10] Liang C, Berant J, Le Q, et al. Neural symbolic machines: Learning semantic parsers on freebase with weak supervision[J]. arXiv preprint arXiv:1611.00020, 2016. [11] Steedman M. A very short introduction to CCG[J]. Unpublished paper. http://www. coqsci. ed. ac. uk/steedman/paper. html, 1996. [12] Liang P. Lambda dependency-based compositional semantics[J]. arXiv preprint arXiv:1309.4408, 2013. [13] Dahl D A, Bates M, Brown M, et al. Expanding the scope of the ATIS task: The ATIS-3 corpus[C]//Proceedings of the workshop on Human Language Technology. Association for Computational Linguistics, 1994: 43-48. [14] Yih S W, Chang M W, He X, et al. Semantic parsing via staged query graph generation: Question answering with knowledge base[J]. 2015. [15]  Yang Y, Chang M W. S-mart: Novel tree-based structured learning algorithms applied to tweet entity linking[J]. arXiv preprint arXiv:1609.08075, 2016. [16] Yu M, Yin W, Hasan K S, et al. Improved neural relation detection for knowledge base question answering[J]. arXiv preprint arXiv:1704.06194, 2017. [17] Luo K, Lin F, Luo X, et al. Knowledge base question answering via encoding of complex query graphs[C]//Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. 2018: 2185-2194. [18] Zhu S, Cheng X, Su S. Knowledge-based question answering by tree-to-sequence learning[J]. Neurocomputing, 2020, 372: 64-72. [19] Wu P, Zhang X, Feng Z. A Survey of Question Answering over Knowledge Base[C]//China Conference on Knowledge Graph and Semantic Computing. Springer, Singapore, 2019: 86-97. [20] Dong L, Lapata M. Language to logical form with neural attention[J]. arXiv preprint arXiv:1601.01280, 2016. [21] Xu K, Wu L, Wang Z, et al. Exploiting rich syntactic information for semantic parsing with graph-to-sequence model[J]. arXiv preprint arXiv:1808.07624, 2018. [22] Hu S, Zou L, Zhang X. A state-transition framework to answer complex questions over knowledge base[C]//Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. 2018: 2098-2108. [23] Cheng J, Reddy S, Saraswat V, et al. Learning structured natural language representations for semantic parsing[J]. arXiv preprint arXiv:1704.08387, 2017. [24] Dyer C, Ballesteros M, Ling W, et al. Transition-based dependency parsing with stack long short-term memory[J]. arXiv preprint arXiv:1505.08075, 2015. [25] Weston J, Chopra S, Bordes A. Memory networks[J]. arXiv preprint arXiv:1410.3916, 2014. [26] Miller A, Fisch A, Dodge J, et al. Key-value memory networks for directly reading documents[J]. arXiv preprint arXiv:1606.03126, 2016. [27] Xu K, Lai Y, Feng Y, et al. Enhancing Key-Value Memory Neural Networks for Knowledge Based Question Answering[C]//Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019: 2937-2947. [28] Yao X, Van Durme B. Information extraction over structured data: Question answering with freebase[C]//Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2014: 956-966. [29] Bordes A, Chopra S, Weston J. Question answering with subgraph embeddings[J]. arXiv preprint arXiv:1406.3676, 2014. [30] Dong L, Wei F, Zhou M, et al. Question answering over freebase with multi-column convolutional neural networks[C]//Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015: 260-269. [31] Hao Y, Zhang Y, Liu K, et al. An end-to-end model for question answering over knowledge base with cross-attention combining global knowledge[C]//Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2017: 221-231. [32] Qu Y, Liu J, Kang L, et al. Question answering over freebase via attentive RNN with similarity matrix based CNN[J]. arXiv preprint arXiv:1804.03317, 2018, 38. [33] Sukhbaatar S, Szlam A, Weston J, et al. Weakly supervised memory networks[J]. CoRR, abs/1503.08895, 2. [34] Bordes A, Usunier N, Chopra S, et al. Large-scale simple question answering with memory networks[J]. arXiv preprint arXiv:1506.02075, 2015. [35]  Jain S. Question answering over knowledge base using factual memory networks[C]//Proceedings of the NAACL Student Research Workshop. 2016: 109-115. [36] Chen Y, Wu L, Zaki M J. Bidirectional attentive memory networks for question answering over knowledge bases[J]. arXiv preprint arXiv:1903.02188, 2019. [37] Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]//Advances in neural information processing systems. 2017: 5998-6008. [38] Reddy S, T_ckstr_m O, Collins M, et al. Transforming dependency structures to logical forms for semantic parsing[J]. Transactions of the Association for Computational Linguistics, 2016, 4: 127-140. [39] Abujabal A, Yahya M, Riedewald M, et al. Automated template generation for question answering over knowledge graphs[C]//Proceedings of the 26th international conference on world wide web. 2017: 1191-1200. [40] Cui W, Xiao Y, Wang H, et al. KBQA: learning question answering over QA corpora and knowledge bases[J]. arXiv preprint arXiv:1903.02419, 2019. [41] Talmor A, Berant J. The web as a knowledge-base for answering complex questions[J]. arXiv preprint arXiv:1803.06643, 2018. [42] Vakulenko S, Fernandez Garcia J D, Polleres A, et al. Message Passing for Complex Question Answering over Knowledge Graphs[C]//Proceedings of the 28th ACM International Conference on Information and Knowledge Management. 2019: 1431-1440. [43] Qiu Y, Wang Y, Jin X, et al. Stepwise Reasoning for Multi-Relation Question Answering over Knowledge Graph with Weak Supervision[C]//Proceedings of the 13th International Conference on Web Search and Data Mining. 2020: 474-482. [44] Petrochuk M, Zettlemoyer L. Simplequestions nearly solved: A new upperbound and baseline approach[J]. arXiv preprint arXiv:1804.08798, 2018. [45] Zhang J, Ye Y, Zhang Y, et al. Multi-Point Semantic Representation for Intent Classification[J]. AAAI. 2020. [46] Jia R, Liang P. Adversarial examples for evaluating reading comprehension systems[J]. arXiv preprint arXiv:1707.07328, 2017. [47] Chakraborty N, Lukovnikov D, Maheshwari G, et al. Introduction to Neural Network based Approaches for Question Answering over Knowledge Graphs[J]. arXiv preprint arXiv:1907.09361, 2019. [48] Ribeiro M T, Singh S, Guestrin C. Semantically equivalent adversarial rules for debugging nlp models[C]//Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2018: 856-865.

https://mp.weixin.qq.com/s/NkwyHsmpZCwsQPiQOMK7zw