Integrated Knowledge Representation Research Articles

Integrating knowledge representation into traffic prediction: a spatial–temporal graph neural network with adaptive fusion features

Various external factors that interfere with traffic flow, such as weather conditions, traffic accidents, incidents, and Points of Interest (POIs), need to be considered in performing traffic forecasting tasks. However, the current research methods encounter difficulties in effectively incorporating these factors with traffic characteristics and efficiently updating them, which leads to a lack of dynamics and interpretability. Moreover, capturing temporal dependence and spatial dependence separately and sequentially can result in issues, such as information loss and model errors. To address these challenges, we present a Knowledge Representation learning-actuated spatial–temporal graph neural network (KR-STGNN) for traffic flow prediction. We combine the knowledge embedding with the traffic features via Gated Feature Fusion Module (GFFM), and dynamically update the traffic features adaptively according to the importance of external factors. To conduct the co-capture of spatial–temporal dependencies, we subsequently propose a spatial–temporal feature synchronous capture module (ST-FSCM) combining dilation causal convolution with GRU. Experimental results on a real-world traffic data set demonstrate that KR-STGNN has superior forecasting performances over diverse prediction horizons, especially for short-term prediction. The ablation and perturbation analysis experiments further validate the effectiveness and robustness of the designed method.

Representation learning for knowledge fusion and reasoning in Cyber–Physical–Social Systems: Survey and perspectives

The digital deep integration of cyber space, physical space and social space facilitates the formation of Cyber–Physical–Social Systems (CPSS). Knowledge empowers CPSS to be capable of solving complex tasks and providing intelligent services. Knowledge representation, fusion, and reasoning are complementary in the knowledge graph and worthy of continuous exploration and in-depth research. Since the concept of knowledge representation learning has been proposed, various models have been developed and demonstrated their superior ability in knowledge fusion and reasoning. In this paper, we investigate the progress and function of representation learning models adopted in knowledge fusion and reasoning, and extract their commonalities and characteristics, which is a perspective that has never been developed before, to provide new ideas for scholars and realize the mutual promotion of knowledge fusion and reasoning. This paper comprehensively reviews classic methods and investigates advanced and emerging works, which are published in authoritative journals or conferences. Furthermore, to overcome the limitations of existing methods, we propose an integrated knowledge representation learning and application framework for CPSS and a series of tensor-based knowledge fusion and reasoning models. Finally, four future directions are discussed in this paper, we analyze the challenges faced by them and suggest promising solutions.

A New Class of Autopoietic and Cognitive Machines

Making computing machines mimic living organisms has captured the imagination of many since the dawn of digital computers. However, today’s artificial intelligence technologies fall short of replicating even the basic autopoietic and cognitive behaviors found in primitive biological systems. According to Charles Darwin, the difference in mind between humans and higher animals, great as it is, certainly is one of degree and not of kind. Autopoiesis refers to the behavior of a system that replicates itself and maintains identity and stability while facing fluctuations caused by external influences. Cognitive behaviors model the system’s state, sense internal and external changes, analyze, predict and take action to mitigate any risk to its functional fulfillment. How did intelligence evolve? what is the relationship between the mind and body? Answers to these questions should guide us to infuse autopoietic and cognitive behaviors into digital machines. In this paper, we show how to use the structural machine to build a cognitive reasoning system that integrates the knowledge from various digital symbolic and sub-symbolic computations. This approach is analogous to how the neocortex repurposed the reptilian brain and paves the path for digital machines to mimic living organisms using an integrated knowledge representation from different sources.

Informed Graph Convolution Networks for Multilingual Short Text Understanding

In the open domain environment, the state-of-the-art models cannot process analyze insufficient training data correctly. We propose an adaptive graph convolution network with informed machine learning for multilingual short text understanding to tackle these problems. Specifically, The prior knowledge to guide the graph neural network to extract sentence topics. We construct category anchor words as prior category keywords, prior category keywords and training data as independent information sources, and prior knowledge participates in the training of graph neural network. Moreover, we integrate the attention mechanism in the training process, so that the model can pay attention to task-related information adaptively. We explain the build blocks and present the integrated knowledge representation. The experimental results on the Multilingual Short Text (MST), THUCNews and AGNews datasets show that our method outperforms most of the existing methods.

Automatic Animal Behavior Analysis: Opportunities for Combining Knowledge Representation with Machine Learning

Computational animal behavior analysis (CABA) is an emerging field which aims to apply AI techniques to support animal behavior analysis. The need for computational approaches which facilitate ‘objectivization’ and quantification of behavioral characteristics of animals is widely acknowledged within several animal-related scientific disciplines. State-of-the-art CABA approaches mainly apply machine learning (ML) techniques, combining it with approaches from computer vision and IoT. In this paper we highlight the potential applications of integrating knowledge representation approaches in the context of ML-based CABA systems, demonstrating the ideas using insights from an ongoing CABA project.

The ethogram project

Computational animal behavior analysis (CABA) is an emerging field which aims to apply AI techniques to support animal behavior analysis. The need for computational approaches which facilitate ‘objectivization’ and quantification of behavioral characteristics of animals is widely acknowledged within several animal-related scientific disciplines. State-of-the-art CABA approaches mainly apply machine learning (ML) techniques, combining it with approaches from computer vision and IoT. In this paper we highlight the potential applications of integrating knowledge representation approaches in the context of ML-based CABA systems, demonstrating the ideas using insights from an ongoing CABA project.

Development of an intelligent environmental knowledge system for sustainable agricultural decision support

The purpose of this research was to develop a knowledge recommendation architecture based on unsupervised machine learning and unified resource description framework (RDF) for integrated environmental sensory data sources. In developing this architecture, which is very useful for agricultural decision support systems, we considered web based large-scale dynamic data mining, contextual knowledge extraction, and integrated knowledge representation methods. Five different environmental data sources were considered to develop and test the proposed knowledge recommendation framework called Intelligent Environmental Knowledgebase (i-EKbase); including Bureau of Meteorology SILO, Australian Water Availability Project, Australian Soil Resource Information System, Australian National Cosmic Ray Soil Moisture Monitoring Facility, and NASA's Moderate Resolution Imaging Spectroradiometer. Unsupervised clustering techniques based on Principal Component Analysis (PCA), Fuzzy-C-Means (FCM) and Self-organizing map (SOM) were used to create a 2D colour knowledge map representing the dynamics of the i-EKbase to provide “prior knowledge” about the integrated knowledgebase. Prior availability of recommendations from the knowledge base could potentially optimize the accessibility and usability issues related to big data sets and minimize the overall application costs. RDF representation has made i-EKbase flexible enough to publish and integrate on the Linked Open Data cloud. This newly developed system was evaluated as an expert agricultural decision support for sustainable water resource management case study in Australia at Tasmania with promising results.

Temporal logic and operation relations based knowledge representation for land cover change web services

Providing land cover spatio-temporal information and geo-computing through web service is a new challenge for supporting global change research, earth system simulation and many other societal benefit areas. This requires an integrated knowledge representation and web implementation of static land cover and change information, as well as the related operations for geo-computing. The temporal logic relations among land cover snapshots and increments were examined with a matrix-based three-step analysis. Twelve temporal logic relations were identified and five basic spatial operations were formalized with set operators, which were all used to develop algorithms for deriving implicit change information. A knowledge representation for land cover change information was then developed based on these temporal logic and operation relations. A prototype web-service system was further implemented based on OWL-DL. Both online access and conversion of land cover spatio-temporal information can be facilitated with such a web service system.

Integrating knowledge representation and quantitative modelling in physiology

A wealth of potentially shareable resources, such as data and models, is being generated through the study of physiology by computational means. Although in principle the resources generated are reusable, in practice, few can currently be shared. A key reason for this disparity stems from the lack of consistent cataloguing and annotation of these resources in a standardised manner. Here, we outline our vision for applying community-based modelling standards in support of an automated integration of models across physiological systems and scales. Two key initiatives, the Physiome Project and the European contribution - the Virtual Phsysiological Human Project, have emerged to support this multiscale model integration, and we focus on the role played by two key components of these frameworks, model encoding and semantic metadata annotation. We present examples of biomedical modelling scenarios (the endocrine effect of atrial natriuretic peptide, and the implications of alcohol and glucose toxicity) to illustrate the role that encoding standards and knowledge representation approaches, such as ontologies, could play in the management, searching and visualisation of physiology models, and thus in providing a rational basis for healthcare decisions and contributing towards realising the goal of of personalized medicine.

Evolution of concept networks and implications for knowledge representation

PurposeThe purpose of this paper is to present descriptive characteristics of the historical development of concept networks. The linguistic principles, mechanisms and motivations behind the evolution of concept networks are discussed. Implications emanating from the idea of the historical development of concept networks are discussed in relation to knowledge representation and organization schemes.Design/methodology/approachNatural language data including both speech and text are analyzed by examining discourse contexts in which a linguistic element such as a polysemy or homonym occurs. Linguistic literature on the historical development of concept networks is reviewed and analyzed.FindingsSemantic sense relations in concept networks can be captured in a systematic and regular manner. The mechanism and impetus behind the process of concept network development suggest that semantic senses in concept networks are closely intertwined with pragmatic contexts and discourse structure. The interrelation and permeability of the semantic senses of concept networks are captured on a continuum scale based on three linguistic parameters: concrete shared semantic sense; discourse and text structure; and contextualized pragmatic information.Research limitations/implicationsResearch findings signify the critical need for linking discourse structure and contextualized pragmatic information to knowledge representation and organization schemes.Originality/valueThe idea of linguistic characteristics, principles, motivation and mechanisms underlying the evolution of concept networks provides theoretical ground for developing a model for integrating knowledge representation and organization schemes with discourse structure and contextualized pragmatic information.

Connectionist modal logic: Representing modalities in neural networks

Modal logics are amongst the most successful applied logical systems. Neural networks were proved to be effective learning systems. In this paper, we propose to combine the strengths of modal logics and neural networks by introducing Connectionist Modal Logics (CML). CML belongs to the domain of neural-symbolic integration, which concerns the application of problem-specific symbolic knowledge within the neurocomputing paradigm. In CML, one may represent, reason or learn modal logics using a neural network. This is achieved by a Modalities Algorithm that translates modal logic programs into neural network ensembles. We show that the translation is sound, i.e. the network ensemble computes a fixed-point meaning of the original modal program, acting as a distributed computational model for modal logic. We also show that the fixed-point computation terminates whenever the modal program is well-behaved. Finally, we validate CML as a computational model for integrated knowledge representation and learning by applying it to a well-known testbed for distributed knowledge representation. This paves the way for a range of applications on integrated knowledge representation and learning, from practical reasoning to evolving multi-agent systems.

An integrated knowledge representation model for the computer-aided conceptual design of mechanisms

An integrated knowledge representation model, namely the topology structure behaviour function (TSBF) model, is presented for the computer-aided conceptual design of mechanisms (MCACD) in this paper. The model covers both qualitative and quantitative knowledge representations of generic mechanisms. A class hierarchy consists of the abstract mechanism, the embryo mechanism, and the concrete mechanism is then proposed for object-oriented modelling. Based on the TSBF model, several reasoning techniques are integrated to achieve a relatively comprehensive environment for MCACD. The corresponding reasoning process is mainly based on a backward chaining of solutions representation and retrieval, a forward chaining of compositional behaviour reasoning with constraint propagation and satisfaction, and a forward chaining of type synthesis. Coarse optimizations for certain mechanisms are also integrated on the quantitative level. The applicability of the new model is demonstrated by the conceptual design of a zigzag mechanism.

Integrating knowledge representation and reasoning in Geographical Information Systems

We propose a formalism and a programming environment in which sophisticated spatio-temporal reasoning can be performed, while keeping the capabilities of manipulating and presenting large amounts of geographical data, typical of commercial Geographical Information Systems (GISs). The spatio-temporal knowledge representation language, named MuTACLP+, is based on constraint logic programming and is integrated via a middleware of commands and translation features with a commercial GIS. The paper presents the language, the architecture of the environment, and a few examples of its use in the field of event planning.

Automated functional design of engineering systems

This paper presents a prototype intelligent system, the knowledge-based functional design automation system (KBFDA) for automating the functional design process of engineering products/systems. An integrated knowledge representation scheme combines rule-based and object-oriented representation methods to represent functions and function related design characteristics in an intelligent design environment. A knowledge-based functional reasoning strategy uses this representation to automatically generate physical behaviors from desired functions or behaviors. The required behaviors are then combined in different configurations to develop a set of potential concept variants that meet the functional requirements and functional constraints given in a design specification. Finally, the variants are ranked according to the degree to which they meet non-functional constraints. The variant with the lowest rank (score) is selected as the best solution. A case study design of a terminal insertion unit is presented to demonstrate the practicality of the proposed approach.

An integrated knowledge representation scheme and query processing mechanism for fault diagnosis in heterogeneous manufacturing environments

An integrated fault diagnosis system was developed for manufacturing complex products, e.g., automobile, that undergo a variety of production processes. The proposed global knowledge representation scheme is designed to consistently express such heterogeneous diagnostic knowledge in terms of the causal relations among relevant attributes across different processes. This scheme is unifying in that it reflects the fact that an effect for one causal relation may in turn be a cause for another along a chain of subsequent processes. In order to unify independent pieces of causal relation, specific attributes of machine-processed products in addition to the usual diagnostic signals such as noise and voltage of the machines. The accompanying probabilistic diagnosis mechanism can then produce an enhanced diagnostic accuracy over other existing schemes as it can simultaneously consider multiple causes of a fault along with their interdependency. The proposed diagnosis mechanism was implemented in a distributed computer environment to identify the functions and procedures required for its actual application.

Integration of multiple knowledge representation for classification problems

This paper describes an approach to integrating various knowledge representations for classification problems. Knowledge representation forms have been analysed. The analysis shows that representation suitability depends on a given situation. Therefore, multiple representation form capability and form conversion capability are both necessary to support developing knowledge bases for wide application areas. A classification problem tool, called HOLON/VP(DT), has been developed with the aim of providing experts with the integrated knowledge representation capability. A knowledge base can be represented in a tabular form, a rule form and a tree form. Form conversion can be accomplished at all times. With this integrated representation, an expert is able to build a knowledge base using the most appropriate form.

Integrated knowledge representation and management in simulation-based design generation

The rising complexity of systems has increased the difficulty in managing knowledge in computer-aided design tools. Conventional representation schemes such as production rules, frames, AND⧸OR trees or semantic networks do not provide sufficient power for managing complex design knowledge. An integrated knowledge representation and management scheme, termed Frames and Rules Associated System Entity Structure (FRASES), is presented for model-based system design applications. A design methodology supported by FRASES is discussed. The methodology combines simulation and artificial intelligence to aid in the design model development and performance evaluation processes. The proposed representation scheme is fundamental to the methodology in that (a) it captures the structure of the system being designed, (b) by organizing complex design knowledge into a hierarchical and entity-based structure, it increases the efficiency of design inference, and (c) it reduces the cost and increases the reliability of a knowledge base. A comprehensive design example illustrating the proposed scheme is also presented.

Integrating discourse pragmatics and propositional knowledge for multilingual natural language processing

We present an integrated knowledge representation system for natural language processing (NLP) whose main distinguishing feature is its emphasis on encoding not only the usual propositional structure of the utterances in the input text, but also capturing an entire complex of nonpropositional — discourse, attitudinal, and other pragmatic — meanings that NL texts always carry. The need for discourse pragmatics, together with generic semantic information, is demonstrated in the context of anaphoric and definite noun phrase resolution for accurate machine translation. The major types of requisite pragmatic knowledge are presented, and an extension of a frame-based formalism developed in the context of the TRANSLATOR system is proposed as a first-pass codification of the integrated knowledge base.

An integrated knowledge representation scheme for expert systems

Abstract The purpose of knowledge representation for an expert system is to specify functions to be performed by the system. In this paper, a knowledge representation scheme which mutually combines procedures, functions, production rules and Horn clauses is outlined. Its knowledge representation model is an imaginary organisation for performing functions of a target system, where a number of members try to solve given problems systematically. Knowledge is distributed to each of the members with considerable modularity. Functional specification of expert systems would be performed with less difficulty.

Mediating the views of databases and database users

The Natural-Language/Deduction group at SRI International has undertaken several large projects integrating knowledge representation, the modeling and use of distributed conventional databases, logical deduction, and natural-language processing. One of the largest projects, LADDER [2], involved accessing data distributed over a computer network by using queries expressed in English. Work with LADDER (and several similar systems) has revealed that: (1) Users wish to talk about data in terms of the enterprise in which the data are to be used. Users do not confine their questions to concepts and terminology covered by the database per se. (2) Users are seldom satisfied with access only to the data in a database. They need to know the KIND of data available (i.e., they want to ask questions about the DB schema), and they expect systems to include information that can be computed from “common knowledge” and information stored explicitly in the database (e.g., if a database records where two ships are, users expect the system to know the distance between them). (3) Users are not satisfied with access to an existing database. They want to tell the system new facts. Some of these are not suitable for storage in conventional databases (e.g., statements involving quantification), and some involve counter factuals (e.g., “Suppose the ship were 100 miles south of its current location...”). (4) Given natural-language access to a DBMS, users expect to interact in natural language with other types of software, too. Moreover, they expect the various underlying software packages to understand one another's results (e.g., User: “Who is the commander of the ship?” System: “Admiral Brown.” User: “Send him a copy of Smith's memo.” The mailer is expected to understand the output from the database).