IndRT-GCNets: Knowledge Reasoning with Independent Recurrent Temporal Graph Convolutional Representations

Due to the structural dependencies among concurrent events in the knowledge graph and the substantial amount of sequential correlation information carried by temporally adjacent events,we propose an Independent Recurrent Temporal Graph Convolution Networks(IndRT-GCNets)framework to efficiently and accurately capture event attribute information.The framework models the knowledge graph sequences to learn the evolutionary represen-tations of entities and relations within each period.Firstly,by utilizing the temporal graph convolution module in the evolutionary representation unit,the framework captures the structural dependency relationships within the knowledge graph in each period.Meanwhile,to achieve better event representation and establish effective correlations,an independent recurrent neural network is employed to implement auto-regressive modeling.Furthermore,static attributes of entities in the entity-relation events are constrained andmerged using a static graph constraint to obtain optimal entity representations.Finally,the evolution of entity and relation representations is utilized to predict events in the next subsequent step.On multiple real-world datasets such as Freebase13(FB13),Freebase 15k(FB15K),WordNet11(WN11),WordNet18(WN18),FB15K-237,WN18RR,YAGO3-10,and Nell-995,the results of multiple evaluation indicators show that our proposed IndRT-GCNets framework outperforms most existing models on knowledge reasoning tasks,which validates the effectiveness and robustness.

基于Graph-LSTMs的双重位置感知方面级情感分类

目前针对用户评论中方面词项情感分类任务的研究大多忽略了依存句法信息,或并未建立依存句法结构与单词之间的联系。为此,提出一种基于Graph-LSTMs的双重位置感知方面级情感分类方法。通过Graph-LSTMs学习词项的上下文语境特征;在双向GRU的输入中拼接具有双重位置信息的位置向量,优化句子情感编码;利用注意力机制捕获关键的情感特征,实现分类。在SemEval2014的两个数据集上的实验结果表明,该模型相比几种基线模型在准确率和Macro-F1这两个指标上提升明显。

The Dynamic-to-Static Conversion of Dynamic Fault Trees Using Stochastic Dependency Graphs and Stochastic Activity Networks

In this paper a new modeling framework for the dependability analysis of complex systems is presented and related to dynamic fault trees (DFTs). The methodology is based on a modular approach: two separate models are used to handle, the fault logic and the stochastic dependencies of the system. Thus, the fault schema, free of any dependency logic, can be easily evaluated, while the dependency schema allows the modeler to design new kind of non-trivial dependencies not easily caught by the traditional holistic methodologies. Moreover, the use of a dependency schema allows building a pure behavioral model that can be used for various kinds of dependability studies. In the paper is shown how to build and integrate the two modular models and convert them in a Stochastic Activity Network. Furthermore, based on the construction of the schema that embeds the stochastic dependencies, the procedure to convert DFTs into static fault trees is shown, allowing the resolution of DFTs in a very efficient way.

Load Dependent Series-Parallel Systems with Common Bus Performance Sharing Mechanism

A series-parallel system was proposed with common bus performance sharing in which the performance and failure rate of the element depended on the load it was carrying. In such a system,the surplus performance of a sub-system can be transmitted to other deficient sub-systems. The transmission capacity of the common bus performance sharing mechanism is a random variable. Effects of load on element performance and failure rate were considered in this paper. A reliability evaluation algorithm based on the universal generating function technique was suggested. Numerical experiments were conducted to illustrate the algorithm.

Graph Convolutional Networks Embedding Textual Structure Information for Relation Extraction

Deep neural network-based relational extraction research has made significant progress in recent years,andit provides data support for many natural language processing downstream tasks such as building knowledgegraph,sentiment analysis and question-answering systems.However,previous studies ignored much unusedstructural information in sentences that could enhance the performance of the relation extraction task.Moreover,most existing dependency-based models utilize self-attention to distinguish the importance of context,whichhardly deals withmultiple-structure information.To efficiently leverage multiple structure information,this paperproposes a dynamic structure attention mechanism model based on textual structure information,which deeplyintegrates word embedding,named entity recognition labels,part of speech,dependency tree and dependency typeinto a graph convolutional network.Specifically,our model extracts text features of different structures from theinput sentence.Textual Structure information Graph Convolutional Networks employs the dynamic structureattention mechanism to learn multi-structure attention,effectively distinguishing important contextual features invarious structural information.In addition,multi-structure weights are carefully designed as amergingmechanismin the different structure attention to dynamically adjust the final attention.This paper combines these featuresand trains a graph convolutional network for relation extraction.We experiment on supervised relation extractiondatasets including SemEval 2010 Task 8,TACRED,TACREV,and Re-TACED,the result significantly outperformsthe previous.

NGDcrm:a numeric graph dependency-based conflict resolution method for knowledge graph

Knowledge graph(KG)conflict resolution is to solve knowledge conflicts problem in the construction of KG.Aiming at the problem of KG conflict resolution,a KG conflict resolution algorithm NGDcrm is proposed,which is a numeric graph dependency-based conflict resolution method.NGDcrm utilizes the dependency graph to perform arithmetic calculation and predicate comparison of numerical entity knowledge in the KG.NGDcrm first uses a parallel segmentation method to segment the KG;then,it extracts the features of the KG according to KG embedding;finally,it uses numerical graph dependencies to detect and correct the wrong facts in the KG based on the extracted features.The experimental results on real data show that NGDcrm is better than the state-of-the-art knowledge conflict resolution method.Among them,the AUC value of NGDcrm on the DBpedia dataset is 15.4%higher than the state-of-the-art method.

System Dependence Graph Construction for Aspect Oriented C++

This paper proposes an extended system dependence graph called AspectSDG to represent control and data dependences for AspeetC＋＋ programs, and presents an approach for the construction of AspectSDG. This approach decomposes aspect-oriented programs into three parts： component codes, aspect codes, and weaving codes. It constructs program dependence graphs （PDGs） for each part, and then connects the PDGs at call sites to form the complete AspectSDG. The AspectSDG can deal with advice precedence correctly, and represent the additional dependences caused by aspect codes. Based on this model, we introduce how to compute a static slice of an AspectC＋ ＋ program.

THE Z SPECIFICATION DEPENDENCE GRAPH

The lack of existing solutions makes it really hard to understand formal specification languages since the application domain for representations is useful for the purpose of carrying out certain software engineering operations such as slicing and the computation of program metrics.A Z specification dependence graph is presented in this letter. It draws on the strengths of a range of earlier works and adapts them, if necessary, to the Z language.

面向交通流量预测的时空Graph-CoordAttention网络

交通预测是城市智能交通系统的一个重要研究组成部分,使人们的出行更加效率和安全。由于复杂的时间和空间依赖性,准确预测交通流量仍然是一个巨大的挑战。近年来,图卷积网络(GCN)在交通预测方面表现出巨大的潜力,但基于GCN的模型往往侧重于单独捕捉时间和空间的依赖性,忽视了时间和空间依赖性之间的动态关联性,不能很好地融合它们。此外,以前的方法使用现实世界的静态交通网络来构建空间邻接矩阵,这可能忽略了动态的空间依赖性。为了克服这些局限性,并提高模型的性能,提出了一种新颖的时空Graph-CoordAttention网络(STGCA)。具体来说,提出了时空同步模块,用来建模不同时刻的时空依赖交融关系。然后,提出了一种动态图学习的方案,基于车流量之间数据关联,挖掘出潜在的图信息。在4个公开的数据集上和现有基线模型进行对比实验,STGCA表现了优异的性能。

Analyzing Ethereum Smart Contract Vulnerabilities at Scale Based on Inter-Contract Dependency

Smart contracts running on public blockchains are permissionless and decentralized,attracting both developers and malicious participants.Ethereum,the world’s largest decentralized application platform on which more than 40 million smart contracts are running,is frequently challenged by smart contract vulnerabilities.What’s worse,since the homogeneity of a wide range of smart contracts and the increase in inter-contract dependencies,a vulnerability in a certain smart contract could affect a large number of other contracts in Ethereum.However,little is known about how vulnerable contracts affect other on-chain contracts and which contracts can be affected.Thus,we first present the contract dependency graph(CDG)to perform a vulnerability analysis for Ethereum smart contracts,where CDG characterizes inter-contract dependencies formed by DELEGATECALL-type internal transaction in Ethereum.Then,three generic definitions of security violations against CDG are given for finding respective potential victim contracts affected by different types of vulnerable contracts.Further,we construct the CDG with 195,247 smart contracts active in the latest blocks of the Ethereum and verify the above security violations against CDG by detecting three representative known vulnerabilities.Compared to previous large-scale vulnerability analysis,our analysis scheme marks potential victim contracts that can be affected by different types of vulnerable contracts,and identify their possible risks based on the type of security violation actually occurring.The analysis results show that the proportion of potential victim contracts reaches 14.7%,far more than that of corresponding vulnerable contracts(less than 0.02%)in CDG.

Adaptive Graph Convolutional Recurrent Neural Networks for System-Level Mobile Traffic Forecasting

Accurate traffic pattern prediction in largescale networks is of great importance for intelligent system management and automatic resource allocation.System-level mobile traffic forecasting has significant challenges due to the tremendous temporal and spatial dynamics introduced by diverse Internet user behaviors and frequent traffic migration.Spatialtemporal graph modeling is an efficient approach for analyzing the spatial relations and temporal trends of mobile traffic in a large system.Previous research may not reflect the optimal dependency by ignoring inter-base station dependency or pre-determining the explicit geological distance as the interrelationship of base stations.To overcome the limitations of graph structure,this study proposes an adaptive graph convolutional network(AGCN)that captures the latent spatial dependency by developing self-adaptive dependency matrices and acquires temporal dependency using recurrent neural networks.Evaluated on two mobile network datasets,the experimental results demonstrate that this method outperforms other baselines and reduces the mean absolute error by 3.7%and 5.6%compared to time-series based approaches.

融合情感语义与句法结构的中文开放域事理图谱构建研究

为解决大规模开放域事理图谱构建过程中缺少标注数据以及事件类型未知导致的限定域事理图谱构建方法难以迁移的问题,利用规则匹配方法高效识别开放域文本中包含的多种事件逻辑关系,融合情感语义与句法结构信息分析提高事件抽取准确性,以更好完成事理图谱的构建任务。首先,总结并扩展因果、顺承、条件、转折等多种逻辑关系抽取模板,并基于规则模板、依存句法信息筛选逻辑关系事件句;其次,创新性地引入情感语义分析方法,在句法结构信息的基础上,通过捕获事件及事件间关系的情感语义精准识别事件类型,进而抽取事件论元;再次,计算语义相似度,进行事件融合,构建<前序事件,事件逻辑关系,后序事件>三元组,得到事件事理图谱,并进一步进行事件泛化以构建抽象事理图谱;最后,以事件发展较完整的“2022年猴痘事件”为数据源,通过实证分析证明开放域事理图谱构建方法可以实现不同类型事件的识别、事件间逻辑关系的揭露,其有效性、可行性得到验证。研究不仅弥补了现有事理图谱构建理论的不足,也为决策支持、事件发展预测等提供有力的数据支持。

面向代码搜索的函数功能多重图嵌入

如何提高异构的自然语言查询输入和高度结构化程序语言源代码的匹配准确度,是代码搜索的一个基本问题.代码特征的准确提取是提高匹配准确度的关键之一.代码语句表达的语义不仅与其本身有关,还与其所处的上下文相关.代码的结构模型为理解代码功能提供了丰富的上下文信息.提出一个基于函数功能多重图嵌入的代码搜索方法.在所提方法中,使用早期融合的策略,将代码语句的数据依赖关系融合到控制流图中,构建函数功能多重图来表示代码.该多重图通过数据依赖关系显式表达控制流图中缺乏的非直接前驱后继节点的依赖关系,增强语句节点的上下文信息.同时,针对多重图的边的异质性,采用关系图卷积网络方法从函数多重图中提取代码的特征.在公开数据集的实验表明,相比现有基于代码文本和结构模型的方法,所提方法的MRR提高5%以上.通过消融实验也表明控制流图较数据依赖图在搜索准确度上贡献较大.

基于胶囊异构图注意力网络的中文表格型数据事实验证

事实验证旨在检查一个文本陈述是否被给定的证据所支持.由于表格结构上具有依赖性、内容上具有隐含性,以表格作为证据的事实验证任务仍面临很多挑战.现有工作或者利用逻辑表达式来解析基于表格证据的陈述,或者设计表格感知神经网络来编码陈述-表格对,以此实现基于表格的事实验证任务.但是,这些方法没有充分利用陈述背后隐含的表格信息,从而导致模型的推理性能下降,并且基于表格证据的中文陈述具有更加复杂的语法和语义,也给模型推理带来更大的困难.为此,提出基于胶囊异构图注意力网络(CapsHAN)的中文表格型数据事实验证方法,所提方法能充分理解陈述的结构和语义,进而挖掘和利用陈述所隐含的表格信息,有效提升基于表格的事实验证任务准确性.具体而言,首先通过对陈述进行依存句法分析和命名实体识别来构建异构图,接着对该图采用异构图注意力网络和胶囊图神经网络进行学习和理解,然后将得到的陈述文本表示与经过编码的表格文本表示进行拼接,最后完成结果的预测.更进一步,针对现有中文表格型事实验证数据集匮乏而难以支持基于表格的事实验证方法性能评价的难题,首先对主流TABFACT和INFOTABS表格事实验证英文数据集进行中文转化,并且专门针对中文表格型数据的特点构建了基于UCL国家标准的数据集UCLDS,该数据集将维基百科信息框作为人工注释的自然语言陈述的证据,并被标记为蕴含、反驳或中立3类.UCLDS在同时支持单表和多表推理方面比传统TABFACT和INFOTABS数据集更胜一筹.在上述3个中文基准数据集上的实验结果表明,所提模型的表现均优于基线模型,证明该模型在基于中文表格的事实验证任务上的优越性.

结合句法增强与图注意力网络的方面级情感分类

方面级情感分类旨在识别给定特定方面文本的情感极性,在本领域中,将图神经网络与句法依赖解析相结合是当下热门的研究方向之一,此类方法通过句法解析捕捉句子中词与词之间的关系,依此构建图结构,输入图神经网络中得到情感极性。若句法解析器出现解析错误,将会对以图为基础的图神经网络模型产生巨大影响。为了增强解析器生成的句法依赖树的解析结果,文中提出了一种句法增强图注意力网络,该网络通过融合多个解析器的解析结果,提高句法依赖解析精度,得到更精准的依赖关系句法图;在图注意力网络中使用密集连接机制捕获更丰富的特征,更适配于增强后的句法图,同时引入方面注意力机制捕获方面语义特征。实验结果验证了句法增强方法的有效性,在3个基准数据集上的分类准确度都有所提高,在方面级情感分析领域具有较好的表现。

基于图常量条件函数依赖的图修复规则发现

数据一致性是数据质量管理的一个重要内容。为了提升图数据一致性,大量关系型数据库中的数据依赖理论被引入到图数据库,包括图函数依赖、图关联规则等。图修复规则是最新提出的一种针对图数据的数据依赖规则,具有强大的修复能力,但目前尚无有效的挖掘算法。为了自动生成图修复规则并提高图数据修复的可靠性,提出一种将图常量条件函数依赖转化为图修复规则的方法(GenGRR)。通过图模式在图中匹配同构子图并映射成节点-属性二维表,从表中相应属性域中抽取错误模式把图常量条件函数依赖转化成图属性值修复规则;删去图模式中常量条件函数依赖RHS对应的节点与相连边生成图属性补充规则。基于最大公共同构子图筛选并验证生成图修复规则的一致性。在多个真实数据集上进行测试,验证相比图常量条件函数直接修复图数据,通过转化生成的图修复规则具有更好的修复效果。

基于语序知识的双通道图卷积网络方面级情感分析

当前大多数基于图卷积网络的方面级情感分析方法利用文本的句法知识、语义知识、情感知识构建文本依赖,但少有研究利用文本语序知识构建文本依赖,导致图卷积网络不能有效地利用文本语序知识引导方面项学习上下文情感信息,从而限制了其性能。针对上述问题,提出基于语序知识的双通道图卷积网络(dual-channel graph convolutional network with word-order knowledge,WKDGCN)模型,该模型由语序图卷积网络(word-order graph convolutional network,WoGCN)和情感知识结合语义知识增强的句法图卷积网络(sentiment and attention-enhanced graph convolutional network,SAGCN)组成。具体地,WoGCN基于文本的语序知识构建图卷积网络,由文本的语序依赖引导方面项特征学习上下文情感信息;SAGCN利用SenticNet中的情感知识结合注意力机制增强句法依赖,利用增强后的句法依赖构建图卷积网络,以此引导方面项特征学习上下文情感信息;最后融合两个图卷积网络的特征用于情感分类。此外,设计了一种权重分配策略,在放大方面项权重的同时保持上下文权重一致,避免错误的计算方面项特征与重要特征的语义相关性。在多个公开数据集上的实验结果表明,所提方法优于对比模型。

基于改进时空图卷积网络的道路行程时间预测模型

为提高道路网行程时间预测精度,综合考虑行程时间的空间依赖性、时间依赖性和天气因素影响,提出了基于属性增强和注意力机制的时空图卷积网络模型.首先,构建属性增强单元,将行程时间和天气信息融合;然后,利用图卷积网络捕获空间依赖性,利用门控递归单元捕获时间依赖性,并利用注意力机制增强模型对特征的学习;最后,利用该模型在真实数据集上对未来15、30、45和60 min的行程时间进行预测.结果表明:预测结果的均方根误差(RMSE)分别为0.0453、0.0456、0.0457和0.0468,与其他模型相比表现更优;考虑了时间、空间和天气因素后,相较于不考虑天气的情况,预测误差降低了约10.3%;相较于不考虑空间依赖性的情况,降低了约24.2%,表明所提模型能更好表达时空依赖性和外部条件影响.

基于双粒度图的文档级关系抽取

文档级关系抽取是指在非结构性文本中抽取实体对之间的关系。针对当前文档级关系抽取方法未能充分利用文档语义信息且难以处理文档的噪声干扰问题,提出一种基于双粒度文档图的关系抽取模型,采用一种新型的构图思路以及降噪方法,分别在句间和句内两个层面进行设计。首先,在句间层面使用修辞语篇关系实体提及关系构建修辞语篇关系图RST-graph,采用异步降噪方式生成粗粒度文档图(CGD-graph),缓解了因实体对的句间关系路径长于句内关系路径造成的结构性误剪枝问题。然后,在句内层面采用依存句法关系对文档中的句子进行解析,构造依存句法树(SDT),增强句内语义信息。最后,将SDT和CGD-graph中存在的公共锚点相连接,构造细粒度文档图(FGD-graph)。实验结果表明,与去噪图推理(DGI)模型相比,该模型的lgn F1值和F1值分别提升了0.40和0.51个百分点,并且在实体对的多标签关系上随着标签数量的增多抽取效果提升较为显著。

融合词法句法信息的方面级情感分析模型

为解决现有方面级情感分析方法缺乏句法约束和词义信息的问题,将句法依存树和知识图谱融合起来对句子编码,提出一种词法句法相结合的图神经网络模型。利用图神经网络分别提取句法依存树中的句法信息和知识图谱中的词法信息,经过位置编码模块和掩码加权模块捕捉重要性更高的单词;将两种特征进行结合获得融合句法词法信息的文本表示,进行情感分类。在3个公开数据集上的实验结果验证了该模型的有效性。