The structure and content of programming knowledge: Disentangling training and language effects in theories of skill development

This article reviews the extensive literature emerging from studies concerned with skill acquisition and the development of knowledge representation in programming. In particular, it focuses upon theories of program comprehension that suggest programming knowledge can be described in terms of stereotypical knowledge structures that can in some way capture programming expertise independently of the programming language used and in isolation from a programmer's specific training experience. An attempt is made to demonstrate why existing views are inappropriate. On the one hand, programs are represented in terms of a variety of formal notations ranging from the quasi‐mathematical to the near textual. It is argued that different languages may lead to different forms of knowledge representation, perhaps emphasizing certain structures at the expense of others or facilitating particular strategies. On the other hand, programmers are typically taught problem‐solving techniques that suggest a strict approach to problem decomposition. Hence, it seems likely that another factor that may mediate the development of knowledge representation, and that has not received significant attention elsewhere, is related to the training experience that programmers typically encounter. In this article, recent empirical studies that have addressed these issues are reviewed, and the implications of these studies for theories of skill acquisition and for knowledge representation are discussed. In conclusion, a more extensive account of knowledge representation in programming is presented that emphasizes training effects and the role played by specific language features in the development of knowledge representation within the programming domain.     

Model formulation as a problem-solving task: Computer-assisted engineering modeling

A central purpose of knowledge acquisition technology is to assist with the formulation of domain models that underlie knowledge systems. In this article we examine the model formulation process itself as a problem-solving task. Drawing from AI research in qualitative reasoning about physical systems, we characterize the model formulation task in terms of the inputs, the reasoning subtasks, and the knowledge needed to perform the problem solving. We describe the elements of a high-level representation of modeling knowledge, and techniques for providing intelligent assistance to the model builder. Applying the results from engineering modeling to knowledge acquisition in general, we identify properties of the representation that facilitate the construction of knowledge systems from libraries of reusable models. © 1993 John Wiley & Sons, Inc.     

An organization of knowledge for problem solving and language comprehension

Plan synthesis and language comprehension, or more generally, the act of discovering how one perception relates to others, are two sides of the same coin, because they both rely on a knowledge of cause and effect—algorithmic knowledge about how to do things and how things work. I will describe a new theory of representation for commonsense algorithmic world knowledge, then show how this knowledge can be organized into larger memory structures, as it has been in a LISP implementation of the theory. The large-scale organization of the memory is based on structures called bypassable causal selection networks. A system of such networks serves to embed thousands of small commonsense algorithm patterns into a larger fabric which is directly usable by both a plan synthesizer and a language comprehender. Because these bypassable networks can adapt to context, so will the plan synthesizer and a language comprehender. I will propose that the model is an approximation to the way humans organize and use algorithmic knowledge, and as such, that it suggests approaches not only to problem solving and language comprehension, but also to learning. I'll describe the commonsense algorithm representation, show how the system synthesizes plans using this knowledge, and trace through the process of language comprehension, illustrating how it threads its way through these algorithmic structures.     

Prolog/Rex-a way to extend Prolog for better knowledgerepresentation

Prolog/Rex represents a powerful amalgamation of the latest techniques for knowledge representation and processing, rich in semantic features that ease the difficult task of encoding heterogeneous knowledge of real-world applications. The Prolog/Rex concept mechanism lets a user represent domain entities in terms of their structural and behavioral properties, including multiple inheritance, arbitrary user-defined relations among entities, annotated values (demons), incomplete knowledge, etc. A flexible rule language helps the knowledge engineer capture human expertise and provide flexible control of the reasoning process. Additional Prolog/Rex strength that cannot be found in any other hybrid language made on top of Prolog is language level support for keeping many potentially contradictory solutions to a problem, allowing possible solutions and their implications to be automatically generated and completely explored before they are committed. The same mechanism is used to model time-states, which are useful in planning and scheduling applications of Prolog/Rex     

复杂问题知识转换的综合集成

复杂问题处理是一项知识密集型任务,在综合集成法的应用中对于相关知识转换过程的把握具有重要意义。文中针对问题的复杂性特点提出一种知识转换过程模型,基于显性知识同隐性知识间的转换,经过认知、外化、集成、内化一系列螺旋上升的动态知识过程获得知识创新。该模型被用于支撑处理复杂问题的方法论——综合集成法,丰富充实问题研讨流程。通过复杂灾害问题实例说明结合知识管理的综合集成法应用。     

Transfer of problem-solving strategy using Covlan

Psychological evidence suggests that humans use visual knowledge and reasoning in solving complex problems. We present Covlan, a visual knowledge representation language for representing visual knowledge and supporting visual reasoning. We describe Galatea, a computer program that uses Covlan for analogical transfer of problem-solving procedures from known analogs to new problems. We present the use of Galatea to model analogical visual problem solving by four human experimental participants, and describe one of the four cases in detail. The Galatea model of human problem solving suggests that problem-solving procedures can be effectively represented with Covlan.     

从经典逻辑知识构建ASP知识库的新方法

回答集程序设计(ASP)是一种主流的非单调知识表示工具。为了能够在利用ASP求解问题过程中使用现有的以经典逻辑表示的知识,给出了一种把以谓词逻辑公式表示的约束型知识和定义型知识转化为ASP程序或知识库的新方法,并以实例说明了其有效性。该方法满足转化后ASP程序的回答集与原公式集的模型具有一一对应关系。在实际应用中,该方法提供了一项从现存的以谓词逻辑为表示语言的知识库,构建以ASP为知识表示语言的非单调知识库的技术。     

融合实体描述与路径信息的知识图谱表示学习模型

知识图谱表示学习方法是将知识图谱中的实体和关系通过特定规则表示成一个多维向量的过程。现有表示学习方法多用于解决单跳知识图谱问答任务,其多跳推理能力无法满足实际需求,为提升多跳推理能力,提出一种融合实体描述与路径信息的知识图谱表示学习模型。首先通过预训练语言模型RoBERTa得到融合实体描述的实体、关系表示学习向量;其次利用OPTransE将知识图谱转化成融入有序关系路径信息的向量。最后构建总能量函数,将针对实体描述和路径信息的向量进行融合。通过实验分析与对比该模型在链路预测任务上与主流知识图谱表示学习模型的性能,验证了该模型的可行性与有效性。     

What kind of thing is a concept?

We discuss several features of concepts used for common knowledge and argue that these features are not superficial, lexical level language-dependent issues, but deep characteristics of the knowledge itself. It is thus necessary to build knowledge representation systems compatible with these characteristics.
We show that the most common suggestions to cope with typicality (e.g., many-valued and (or) nonmonotonic systems) fail to capture entirely this phenomenon. As for the other features, no serious attempt has been made yet, and we only propose tentative elements for a solution.
The main idea is to decouple the notion of concept from the notion of basic element (predicate, node), and to represent a concept by an open-ended family of entities of the system. Each entity conveys a possible interpretation of the concept, and interpretations are ordered, according to their "depth." An example illustrating the main features of this scheme is provided.     

面向链接预测的知识图谱表示学习方法综述

作为人工智能的重要基石, 知识图谱能够从互联网海量数据中抽取并表达先验知识, 极大程度解决了智能系统认知决策可解释性差的瓶颈问题, 对智能系统的构建与应用起关键作用. 随着知识图谱技术应用的不断深化, 旨在解决图谱欠完整性问题的知识图谱补全工作迫在眉睫. 链接预测是针对知识图谱中缺失的实体与关系进行预测的任务, 是知识图谱构建与补全中不可或缺的一环. 要充分挖掘知识图谱中的隐藏关系, 利用海量的实体与关系进行计算, 就需要将符号化表示的信息转换为数值形式, 即进行知识图谱表示学习. 基于此, 面向链接预测的知识图谱表示学习成为知识图谱领域的研究热点. 从链接预测与表示学习的基本概念出发, 系统性地介绍面向链接预测的知识图谱表示学习方法最新研究进展. 具体从知识表示形式、算法建模方式两种维度对研究进展进行详细论述. 以知识表示形式的发展历程为线索, 分别介绍二元关系、多元关系和超关系知识表示形式下链接预测任务的数学建模. 基于表示学习建模方式, 将现有方法细化为4类模型: 平移距离模型、张量分解模型、传统神经网络模型和图神经网络模型, 并详细描述每类模型的实现方式与解决不同关系元数链接预测任务的代表模型. 在介绍链接预测的常用的数据集与评判标准基础上, 分别对比分析二元关系、多元关系和超关系3类知识表示形式下, 4类知识表示学习模型的链接预测效果, 并从模型优化、知识表示形式和问题作用域3个方面展望未来发展趋势.     

Toward a model of representation changes

This paper presents the first steps in the development of a computer model of the process of changing representations in problem solving. The task of discovering representations that yield efficient solution strategies for problems is viewed as heuristic search in the space of representations. Two dimensions of this representation space are information structure and information quantity. Changes of representation are characterized as isomorphisms and homomorphisms, corresponding to changes of information structure and information quantity, respectively. A language for expressing representations is given. Also, a language for describing representation transformations and an interpreter for applying the transformations to representations has been developed. In addition, transformations can be automatically inverted and composed to generate new transformations. Among the example problems used to illustrate and support this model are tic-tac-toe, integer arithmetic, the Tower of Hanoi problem, the arrow puzzle, the five puzzle, the mutilated checkerboard problem, and floor plan design. The system has also been used to generate some new NP-complete problems.     

Unified problem modeling language for knowledge engineering of complex systems

The 90s has seen the emergence of hybrid configurations of four most commonly used intelligent methodologies, namely, symbolic knowledge based systems (e.g. expert systems), artificial neural networks, fuzzy systems and genetic algorithms. These hybrid configurations are used for different problem solving tasks/situations. In this paper we describe unified problem modeling language at two different levels, the task structure level for knowledge engineering of complex data intensive domains, and the computational level of the task level hybrid architecture. Among other aspects, the unified problem modeling language considers various intelligent methodologies and their hybrid configurations as technological primitives used to accomplish various tasks defined at the task structure level. The unified problem modeling language is defined in the form of five problem solving adapters. The problem solving adapters outline the goals, tasks, percepts/inputs, and hard and soft computing methods for modeling complex problems. The task structure level has been applied in modeling several applications in e-commerce, image processing, diagnosis, and other complex, time critical, and data intensive domains. We also define a layered intelligent multi-agent, operating system processes, intelligent technologies with the task structure level associative hybrid architecture. The layered architecture also facilitates component based software modeling process.Work Supported by VPAC grant no EPPNLA002.2001     

Domains and context: First steps towards managing diversity in knowledge

Despite the progress made, one of the main barriers towards the use of semantics is the lack of background knowledge. Dealing with this problem has turned out to be a very difficult task because on the one hand the background knowledge should be very large and virtually unbound and, on the other hand, it should be context sensitive and able to capture the diversity of the world, for instance in terms of language and knowledge. Our proposed solution consists in addressing the problem in three steps: (1) create an extensible diversity-aware knowledge base providing a continuously growing quantity of properly organized knowledge; (2) given the problem, build at run-time the proper context within which perform the reasoning; (3) solve the problem. Our work is based on two key ideas. The first is that of using domains, i.e. a general semantic-aware methodology and technique for structuring the background knowledge. The second is that of building the context of reasoning by a suitable combination of domains. Our goal in this paper is to introduce the overall approach, show how it can be applied to an important use case, i.e. the matching of classifications, and describe our first steps towards the construction of a large scale diversity-aware knowledge base.     

Impact of Knowledge Representations on Problem-Oriented Learning in Online Environments

The objective of this study was to investigate the impact of knowledge representations on problem-oriented learning in online learning environments. The study compared the impact of knowledge map representation with traditional hierarchical representation with regard to learning memory and problem-solving performance. Twenty-nine students participated in an experiment in which they studied online materials with the goal of solving two programming problems (simple and complex). It was found that participants who used the hierarchical representation read in the depth-first sequence, whereas participants who used the knowledge map representation read in a sequence reflecting the system running mechanism implied by the graphical representation. In addition, participants who used the knowledge map representation had better memory of the learning content, especially about relations between knowledge nodes. When solving the complex problem, participants who used the knowledge map representation made a deeper analysis of the problem and had better problem-solving performance. These results were not significant in the simple problem-solving task.     

基于共享BERT和门控多任务学习的事件检测方法

事件检测任务的目标是从文本中自动获取结构化的事件信息。目前基于表示学习的神经事件检测方法能够有效利用潜在语义信息,但人工标注数据集的语义知识含量有限,制约了神经网络模型的认知广度。相对地,多任务表示学习框架,有助于模型同时学习不同任务场景中的语义知识,从而提升其认知广度。BERT预训练模型得益于大规模语言资源的充沛语义信息,具有高适应性(适应不同任务)的语义编码能力。因此,该文提出了一种基于BERT的多任务事件检测模型。该方法将BERT已经包含的语义知识作为基础,进一步提升多任务模型的表示、学习和语义感知能力。实验表明,该方法有效提高了事件检测的综合性能,其在ACE2005语料集上事件分类的F₁值达到了76.7%。此外,该文在实验部分对多任务模型的训练过程进行了详解,从可解释性的层面分析了多任务架构对事件检测过程的影响。     

一个实体关系与事件抽取的通用模型

信息提取的目的是从自然语言文件中找到具体信息,现有研究在信息抽取的实体关系和事件抽取任务中仅解决事件论元重叠和实体关系重叠的问题,未考虑两个任务共有的角色重叠问题,导致抽取结果准确率降低。提出一个两阶段的通用模型用于完成实体关系抽取和事件抽取子任务。基于预训练语言模型RoBERTa的共享特征表示,分别对实体关系/事件类型和实体关系/事件论元进行预测。将传统抽取触发词任务转化为多标签抽取事件类型任务,利用多尺度神经网络进一步提取文本特征。在此基础上,通过抽取文本相关类型的事件论元,根据论元角色的重要性对损失函数重新加权,解决数据不平衡、实体关系抽取和事件抽取中共同存在论元角色重叠的问题。在千言数据集中事件抽取和关系抽取任务测试集上的实验验证了该模型的有效性,结果表明,该模型的F1值分别为83.1%和75.3%。     

Knowledge-intensive natural language generation

The development of natural language interfaces to artificial intelligence systems is dependent on the representation of knowledge. A major impediment to building such systems has been the difficulty in adding sufficient linguistic and conceptual knowledge to extend and adapt their capabilities. This difficulty has been apparent in systems that perform the task of language production, i.e., the generation of natural language output to satisfy the communicative requirements of a system.A uniform, parsimonious representation of knowledge about language can increase extensibility and efficiency as well as simplify the generation process. The Ace framework applies knowledge representation fundamentals to the task of encoding knowledge about language. Within this framework, linguistic and conceptual knowledge are organized into hierarchies, and structured associations are used to join knowledge structures that are metaphorically related or otherwise used in linguistic expression. These structured associations permit specialized linguistic knowledge to derive partially from more abstract knowledge, facilitating the use of abstractions in generating specialized phrases. A simple generator called KING (Knowledge INtensive Generator) uses an Ace knowledge base to produce utterances from their conceptual representation.     

A diagnostic environment for automaton networks

Automated diagnosis of communicating‐automaton networks (CANs) is a complex task, which is typically faced by model‐based reasoning, where the behavior of the network is reconstructed based on its observation. This task may take advantage of knowledge‐compilation techniques, where a large amount of reasoning is anticipated off‐line (when the diagnostic process is not active), by simulating the behavior of the network and by constructing suitable data structures embedding diagnostic information. This (general‐purpose) compiled knowledge is exploited on‐line (when the diagnostic process becomes active), so as to generate the solution to the problem. Additional reusable (special‐purpose) compiled knowledge is generated on‐line when solving new problems. A software environment for the diagnosis of CANs has been developed in the C programming language with the support of the PostgreSQL relational database management system, under the Linux operating system. It supports the modeling and preprocessing of CANs as well as the solution of diagnostic problems, including on‐line knowledge compilation. The environment has been tested through a variety of experiments. Results are encouraging and provide a valuable feedback for further work. Copyright © 2006 John Wiley & Sons, Ltd.     

Logical reasoning in natural language: It is all about knowledge

A formal, computational, semantically clean representation of natural language is presented. This representation captures the fact that logical inferences in natural language crucially depend on the semantic relation of entailment between sentential constituents such as determiner, noun, adjective, adverb, preposition, and verb phrases.The representation parallels natural language in that it accounts for human intuition about entailment of sentences, it preserves its structure, it reflects the semantics of different syntactic categories, it simulates conjunction, disjunction, and negation in natural language by computable operations with provable mathematical properties, and it allows one to represent coordination on different syntactic levels.The representation demonstrates that Boolean semantics of natural language can be successfully modeled in terms of representation and inference by knowledge representation formalisms with Boolean semantics. A novel approach to the problem of automatic inferencing in natural language is addressed. The algorithm for updating a computer knowledge base and reasoning with explicit negative, disjunctive, and conjunctive information based on computing subsumption relation between the representations of the appropriate sentential constituents is discussed with examples.     

Logical form generation as abduction: Part I. Representation of linguistic concepts

For some time, researchers have become increasingly aware that some aspects of natural language processing can be viewed as abductive inference. This article describes knowledge representation in dual-route parsimonious covering theory, based on an existing diagnostic abductive inference model, extended to address issues specific to logic form generation. the two routes of covering deal with syntactic and semantic aspects of language, and are integrated by attributing both syntactic and semantic facets to each “open class” concept. Such extensions reflect some fundamental differences between the two task domains. the syntactic aspect of covering is described to show the differences, and some interesting properties are established. the semantic associations are characterized in terms of how they can be used in an abductive model. A major significance of this work is that it paves the way for a nondeductive inference method for word sense disambiguation and logical form generation, exploiting the associative linguistic knowledge. This approach sharply contrasts with others, where knowledge has usually been laboriously encoded into pattern-action rules that are hard to modify. Further, this work represents yet another application for the general principle of parsimonious covering. © 1994 John Wiley & Sons, Inc.