基于概念层次网络的知识表示与本体建模

知识表示是自然语言理解的重要基础。知识表示不统一、语义信息无法系统化利用是目前存在的亟待解决的问题。要解决这个问题,就要解决语义知识表示的问题。该文基于概念层次网络,描述了词语、句子和篇章层面的语义知识表示方法。基于文中描述的词汇层面的表示方法,构建了一个多语言本体知识库。该知识库的知识表示方法不仅可以为知识表示理论研究提供基础,还可以为自然语言处理相关领域的应用提供资源支持。     

基于知识图谱的案件特征增强法律判决预测

现有基于知识图谱的法律判决预测方法重点关注案件的要素实体和关系,不能充分地获取案件的特征信息。针对该问题,提出了一种增强案件特征融合的知识图谱法律判决预测方法。首先,该方法利用双向门控循环神经网络挖掘事实描述文本深层次的因果、时序等全文语义特征信息。然后通过知识图谱向量空间中案例间相似度注意力计算学习类案特征表示。最后,融合特征信息和知识图谱的结构化知识,丰富实体和关系在案件事实文本中的语义特征表示,实现法律判决链路预测任务。在危险驾驶罪和盗窃罪两类罪名数据集上的实验结果显示,该方法在MRR、Hit@1两个关键评价指标上与当前表现最好的链路预测模型相比提升了1.5%左右,Hit@3和Hit@10等指标也均有提升,验证了案件特征增强融合能补充法律知识图谱中缺失的案件特征信息并提高预测的效果。     

A crowdsourcing approach to building a legal ontology from text

This article focuses on the problems of application of artificial intelligence to represent legal knowledge. The volume of legal knowledge used in practice is unusually large, and therefore the ontological knowledge representation is proposed to be used for semantic analysis, presentation and use of common vocabulary, and knowledge integration of problem domain. At the same time some features of legal knowledge representation in Ukraine have been taken into account. The software package has been developed to work with the ontology. The main features of the program complex, which has a Web-based interface and supports multi-user filling of the knowledge base, have been described. The crowdsourcing method is due to be used for filling the knowledge base of legal information. The success of this method is explained by the self-organization principle of information. However, as a result of such collective work a number of errors are identified, which are distributed throughout the structure of the ontology. The results of application of this program complex are discussed in the end of the article and the ways of improvement of the considered technique are planned.     

基于层次结构的生成式自动文摘

基于编码器—解码器架构的序列到序列学习模型是近年来主流的生成式文摘方法。但是,传统的编码器尚不能有效地对长文档进行语义编码,并且只能学习线性链结构的信息, 忽视了文档具有的层次结构。而文档的层次结构(字—句—文档)有助于自动文摘系统更加准确地判断文档内不同结构单元的语义信息和重要程度。为了使编码器能够获取文档的层次结构信息,该文根据文档的层次结构对文档进行编码: 首先构建字级语义表示,然后由字级语义表示构建句级语义表示。另外,该文还提出了一种语义融合单元来对输入文档不同层次的语义信息进行融合,作为最终的文档表示提供给编码器生成摘要。实验结果表明,在加入该文提出的层次文档阅读器与语义融合单元后,系统性能在 ROUGE 评价指标上有显著提高。     

基于本体的领域知识重用方法研究*

由于已有知识表示和本体知识表示并不在同一逻辑体系基础上,从知识表示层面提出了一种基于知识等价映射的知识重用方法解决对已有异构知识的重用问题。通过语义等价提取、在本体指导下的一致性处理以及语义映射解决了本体构建过程中对已有知识的自动重用问题。实验表明,该方法易于实现且具有较高的精确性、可行性和有效性。     

融合实体描述信息和邻居节点特征的知识表示学习方法

知识图谱表示学习旨在将实体和关系映射到一个低维稠密的向量空间中。现有的大多数相关模型更注重于学习三元组的结构特征,忽略了三元组内的实体关系的语义信息特征和三元组外的实体描述信息特征,因此知识表达能力较差。针对以上问题,提出了一种融合多源信息的知识表示学习模型BAGAT。首先,结合知识图谱特征来构造三元组实体目标节点和邻居节点,并使用图注意力网络（GAT）聚合三元组结构的语义信息表示;然后,使用BERT词向量模型对实体描述信息进行嵌入表示;最后,将两种表示方法映射到同一个向量空间中进行联合知识表示学习。实验结果表明,BAGAT性能较其他模型有较大提升,在公共数据集FB15K-237链接预测任务的Hits@1与Hits@10指标上,与翻译模型TransE相比分别提升了25.9个百分点和22.0个百分点,与图神经网络模型KBGAT相比分别提升了1.8个百分点和3.5个百分点。可见,融合实体描述信息和三元组结构语义信息的多源信息表示方法可以获得更强的表示学习能力。     

Prajna: Adding Automated Reasoning to the Visual- Analysis Process

Applications and systems can represent knowledge in various ways. Graphic displays might help a data analyst infer new information through interactive visualizations. Knowledge represented as a collection of facts can be used for automatic inference, although it might be represented or stored in various archives, such as databases or formatted files. Developers who create applications for knowledge representation frequently must contend with not only data challenges but also challenges caused by a wide variety of software toolkits, architectures, and standards for knowledge representation. To overcome these obstacles, Vision Systems & Technology, Inc. initiated the Prajna project. The result was a Java toolkit designed to provide various capabilities for visualization, knowledge representation, geographic displays, semantic reasoning, and data fusion. This article is part of a special issue on knowledge-assisted visualization.     

融合事件信息的中文问答系统问题语义表征

复杂类问题理解是中文问答系统研究的难点,基于组块的问句分析方法将整个问句转化为若干组块,降低了问句分析的难度和复杂性。针对以含有事件(动作)信息的复杂类问题,提出基于语义组块的中文问答系统问题语义表征模型,采用语义组块的思想将问题的语义成分定义为疑问焦点块、问题主题块和问题事件块三个语义组块,对问句中的事件语义信息,建立了问题事件语义结构,将一个问句表征为一个基于语义组块的问题语义表征结构,用于问答系统的问题理解。通过序列标注学习方法实现问题语义表征中语义组块自动标注。实验结果表明:问题语义组块标注效果较好,问题语义表征模型获取了问题的关键语义信息,为语义层面上的问题理解提供基础。     

融合语义信息的矩阵分解词向量学习模型

词向量在自然语言处理中起着重要的作用,近年来受到越来越多研究者的关注。然而,传统词向量学习方法往往依赖于大量未经标注的文本语料库,却忽略了单词的语义信息如单词间的语义关系。为了充分利用已有领域知识库（包含丰富的词语义信息）,文中提出一种融合语义信息的词向量学习方法（KbEMF）,该方法在矩阵分解学习词向量的模型上加入领域知识约束项,使得拥有强语义关系的词对获得的词向量相对近似。在实际数据上进行的单词类比推理任务和单词相似度量任务结果表明,KbEMF比已有模型具有明显的性能提升。     

A Methodology to Create Legal Ontologies in a Logic Programming Based Web Information Retrieval System

Web legal information retrieval systems need the capability to reason with the knowledge modeled by legal ontologies. Using this knowledge it is possible to represent and to make inferences about the semantic content of legal documents. In this paper a methodology for applying NLP techniques to automatically create a legal ontology is proposed. The ontology is defined in the OWL semantic web language and it is used in a logic programming framework, EVOLP+ISCO, to allow users to query the semantic content of the documents. ISCO allows an easy and efficient integration of declarative, object-oriented and constraint-based programming techniques with the capability to create connections with external databases. EVOLP is a dynamic logic programming framework allowing the definition of rules for actions and events. An application of the proposed methodology to the legal web information retrieval system of the Portuguese Attorney General’s Office is described.     

A graph-based approach for extracting terminological properties from information sources with heterogeneous formats

The problem of handling both the integration and the cooperation of a large number of information sources characterised by heterogeneous representation formats is a challenging issue. In this context, a central role can be played by the knowledge about the semantic relationships holding between concepts belonging to different information sources (intersource properties). In this paper, we propose a semiautomatic approach for extracting two kinds of intersource properties, namely synonymies and homonymies, from heterogeneous information sources. In order to carry out the extraction task, we introduce both a conceptual model, for representing involved sources, and a metrics, for measuring the strength of the semantic relationships holding among concepts represented within the same source.     

Salient object detection using biogeography-based optimization to combine features

In recent studies, ontology related concepts have been introduced into FIPA ACL content language to convey information for agent communication. However, these works have only applied ontology-based knowledge representation in communication message and then demonstrated the advantage of this association. In fact, although ontology can represent semantic implications needed for decidable reasoning support, it has no mechanism for defining complex rule-based representation to support inference. The motivation of this study is to address this issue by developing a semantic-based infrastructure to integrate Semantic Web technologies into ACL message contents. This semantic-based infrastructure defines two different semantic frameworks: the three-tier knowledge representation framework for message content and the Multi-layer Ontology Architecture for content language. The former is developed based on Semantic Web stack to support ontology-based reasoning and rule-based inference. The latter is adopted to develop a Lightweight Ontology-based Content Language (LOCL) to describe agent communication messages in an unambiguous and computer-interpretable way Jena reasoner is used in an application scenario that exploits agent communication with LOCL as content language, OWL as ontology language, and SWRL as rule language to demonstrate the feasibility of the proposed infrastructure.     

Some ideas on semantic memory in automatic learning of heuristics

This paper is concerned with the problem of computer semantic memory, i. e. with the problem of representing general knowledge about, a given world. The sematic memory issue is raised in the context of machiue learning of heuristics and the connection with the problem of machine representation of knowledge is emphasized. The guidelines for the implementation of a sematic memory are presented and attention is given to the design of the processes for storing and retrieving information. The problem of knowledge representation is tackled in its, general form, so that the proposed semantic memory may be of interest also in other fields, like Natural Lauguage Understanding, Question Answering, Theorem Proving. The research described in this paper was carried out wile the author was visiting, the Department of Computer Science, Carnegie Mellon University, Pittsburgh, Pa. U. S. A. The research has been supported by the Advanced Research Projects Agency of the Office of the U. S. Secretary of Defense (F 44620-70-C-0107), monitored by the Air Force Office of Scientific Reseach, and in part by the National Research Council Istituto di Elaborazione dell'informazione, Pisa Italy.     

一种面向云制造的语义BOM知识集成框架

为克服传统BOM信息集成中存在语义知识集成的缺陷, 支持BOM知识在云制造企业间无缝传递、共享和重用, 提出一种面向云制造的语义BOM知识集成框架。该集成框架由语义BOM知识表示、知识映射、知识服务三个模块构成。知识表示模块融合本体论, 用owl DL进行语义BOM的形式化知识表示, 构建语义BOM知识本体模型; 知识映射模块基于语义BOM多视图映射方法, 形成云制造企业所需语义BOM视图; 知识服务模块以Web service为媒介, 将各语义BOM视图封装成知识服务, 通过服务注册、绑定、发现、组合等手段, 实现BOM知识的共享和重用。该框架能支持云制造企业按需在知识重用基础上快速重构BOM知识模型。     

语义万维网的概念、方法及应用

近两年来,语义万维网(semanticweb)的研究逐渐引起了知识表示、逻辑编程、信息系统集成、web开发等各个领域的广泛关注。笔者在研究万维网环境下的领域知识表示及语义共享模式的过程中,阅读了大量有关语义万维网的文献资料,认为,语义万维网的研究将对传统web上信息的发布、存储和处理方式产生一场变革,但是语义万维网的概念、思想和方法还处于形成阶段,国内少有综述性的文献,对语义万维网及其相关技术的认识比较模糊。该文分析了语义万维网的起源、概念、技术框架,总结了语义万维网及相关工具的现状,并讨论了语义万维网技术在智能信息检索、企业间数据交换、知识管理以及万维网服务中的应用。     

Intelligent non-visual navigation of complex HTML structures

This paper provides an overview of a project aimed at using knowledge-based technology to improve accessibility of the Web for visually impaired users. The focus is on the multi-dimensional components of Web pages (tables and frames); our cognitive studies demonstrate that spatial information is essential in comprehending tabular data, and this aspect has been largely overlooked in the existing literature. Our approach addresses these issues by using explicit representations of the navigational semantics of the documents and using a domain-specific language to query the semantic representation and derive navigation strategies. Navigational knowledge is explicitly generated and associated to the tabular and multi-dimensional HTML structures of documents. This semantic representation provides to the blind user an abstract representation of the layout of the document; the user is then allowed to issue commands from the domain-specific language to access and traverse the document according to its abstract layout. Published online: 6 November 2002     

融合案件要素的相似案例匹配

相似案例匹配是智慧司法中的重要任务,其通过对比两篇案例的语义内容判别二者的相似程度,能够应用于类案检索、类案类判等。相对于普通文本,法律文书不仅篇幅更长,文本之间的区别也更微妙,传统深度匹配模型难以取得理想效果。为了解决上述问题,该文根据文书描写规律截取文书文本,并提出一种融合案件要素的方法来提高相似案件的匹配性能。具体来说,该文以民间借贷案件为应用场景,首先基于法律知识制定了6种民间借贷案件要素,利用正则表达式从法律文书中抽取案件要素,并形成词独热形式的案件要素表征;然后,对法律文本倒序截取,并通过BERT编码得到法律文本表征,解决法律文本的长距离依赖问题;接着使用线性网络融合法律文本表征与案件要素表征,并使用BiLSTM对融合的表征进行高维度化表示;最后通过孪生网络框架构建向量表征相似性矩阵,通过语义交互与向量池化进行最终的相似度判断。实验结果表明,该文模型能有效处理长文本并建模法律文本的细微差异,在CAIL2019-SCM公共数据集上优于基线模型。     

基于概念图的汉语语义计算的研究与实现

中文信息处理的发展迫切需要加强汉语语义理论的研究,尤其是汉语语义表示形式和语义计算的研究。针对目前汉语语义计算方法的计算结果并不准确的问题,提出了一种基于概念图的汉语语义计算方法。该方法以“知网”为语义知识资源,以概念图为知识表示方法,把自然语言文本转化为概念图,通过概念图的匹配实现语义计算,以改善语义计算的效果。实验结果表明该方法对汉语语义计算是有效的。     

基于语义Web上知识表示的研究及其应用

随着对语义研究的深入,人们越来越关注在W EB上信息内容的表示问题,通过分析语义W EB上知识表示的特点后指出,以RDF为基础的知识表示语言可较好地实现语义W EB的知识表示。最后通过语义W EB上的语义检索应用给予了说明。