Open Semantic Technologies for Intelligent System

Artificial intelligence (AI) is the future of computer technologies and at present it has achieved great progress in different areas. In this paper, we study the general AI problem from a standardization point of view, and introduce a semantic interoperability for intelligent computer systems. We propose a standard for the interior semantic representation of knowledge in the memory of an intelligent computer system, which is called the SC-code (Semantic Code). Integration of various types of knowledge is performed due to hybrid knowledge base models. This model includes a hierarchical set of top-level ontologies that provide semantic compatibility of various types of knowledge and permits to integrate facts, specifications of various objects, logical statements, events, situations, programs and algorithms, processes, problem formulations, domain models, ontologies, and so on. A variant of the presentation of the AI standard based on the semantic representation of knowledge, in the form of a part of the knowledge base of the Intelligent Computer Metasystem (IMS.ostis) is proposed.

This paper provides an overview of the current state in the development of artificial intelligence (AI) systems. It is discussed that an ideal AI system should imitate (support the development) of all the characteristics (cognitive procedures) of natural intelligence. Moreover, not all characteristics of natural intelligence can be automated; some of them can be realized only using human-computer interactions. An important property of AI systems is the integration of cognitive procedures among themselves. Thus, the AI system should support the implementation of a subset (as large as possible) of natural intelligence procedures. Moreover, all procedures should be understandable to a person (described in his or her system of concepts), and the connection between them should be “seamless” way to ensure their integration. The paper identifies and substantiates the requirements for tools for creating AI systems. It is proposed to evaluate the AI system by the number of cognitive procedures that it implements. At the same time, given the complexity of tools, its developers need to develop a set of requirements for compatibility and integration of tools and AI systems of various types.

This paper considers knowledge convergation problems for different types of intelligence systems with cognitive-based decision-making. The basis is semantic interoperability of knowledge bases, all of which are managed by knowledge metabase of subject domain or several adjacent domains. The concept of using an information system with various intelligent decision support systems provides for the formation of knowledge base rules based on synonyms. To make and to justify decisions in convergent intelligent systems, it is proposed to use two and 3-simplex prisms. 2-simplex prism is more appropriate if analyzed indicators are changing dynamically.

The paper deals with the knowledge discovery problem in subject areas with open empirical data, where formal means are absent and the procedures for the theories’ formation are heuristic. The inherent challenges of the exact epistemology in such sciences are represented. Approaches to solving these problems by means of the JSM Method of automated support for research are described. We present the intelligent system JSM Socio implementing the JSM Method. JSM Socio automatically reproduces an imitation of the natural (rational) intelligence’ abilities to solve various problems of sociological data analysis. We have studied various forms of constructive social activities by the JSM Socio, the results obtained are presented. The JSM Method of automated support for research proved to be the knowledge discovery tool.

In this paper, we propose an approach to automating the processes of creating, developing and applying standards based on OSTIS Technology. The problems of modern approaches to the development, maintenance and application of standards are considered in detail, special attention is paid to standards in the field of Industry 4.0, such as ISA-88 and ISA-95, their role in the context of Industry 4.0 and problems specific to standards in this field are considered. The paper proposes an approach to the development of standards based on the ontological approach and involving the transformation of the standard into a knowledge base developed by a distributed team of developers directly in the process of its use. It is proposed to use OSTIS Technology as the basis for building this kind of system. We consider a prototype information system for employees of a batch production enterprise that implements the proposed approach, as well as examples of the integration of such a system with production systems.

The paper presents the implementation of an intelligent decision support system (IDSS) to solve a real manufacturing problem at JSC “Savushkin Product”. The proposed system is intended to control the quality of product labeling, based on neuro-symbolic artificial intelligence, namely integrating deep neural networks and semantic models. The system perform localization and recognition of images from a high-speed video stream and is based on several deep neural networks. Semantic networks fulfill intelligent processing of recognition results in order to generate final decision as regards the state of the production conveyor. We demonstrate the performance of the proposed technique in the real production process. The main contribution of this paper is a novel view at the creation of a real intelligent decision support system, which combines bio inspired approach, namely neural networks and conventional technique, based on a knowledge base.

In this paper principles of decision-making systems construction are considered. An approach to image analysis based on semantic model is proposed and studied. The results show an improvement in processing speed and image captioning quality based on Visual Genome dataset.

The paper considers the approach to building a personal assistant based on the open semantic technology (OSTIS). The key idea of our approach is the transition from processing a message in speech form to a formalized representation of the meaning in a knowledge base with the least number of intermediate stages. Knowledge base is built on the basis of a semantic network implemented using OSTIS technology. In this case, many tasks of the syntactic, semantic and pragmatic levels of natural language processing, for example, such as recognition of named entities (NEM) and definition of a part of speech (POS), dialogue management (intentions identifications and directives forming), can be performed directly in the knowledge base of the intelligent assistant. This will make it possible to effectively solve such problems as managing the global and local context of the dialogue, resolving linguistic phenomena such as anaphores, homonymy and eleptical phrases, correctly formulating answers to questions that are complex in structure and content, posed by the user during the dialogue.

The natural language user interface is a subclass of user interfaces that allows user and system to communicate using natural language. It is the development direction of the user interface of the intelligent system. The key technology for implementation of natural language user interface is the computer processing of natural language text. Due to the diversity and complexity of natural language, its understanding hasn’t completely achieved yet. By comparing Chinese language with other European languages, this article describes the characteristics of Chinese language and the difficulties in Chinese language processing. After an analysis of current mainstream natural language processing methods, it was shown that the knowledge base plays an important role in the natural language processing model. The knowledge base is the basis for natural language processing. This article proposes a method of computer processing of Chinese language text based on Chinese linguistic ontology and domain ontologies. The ontologies are used to build a unified semantic model of Chinese linguistic knowledge and domain knowledge for the processing of Chinese language text. In this way the Chinese linguistic knowledge is integrated in the Chinese language processing model, the application of Chinese linguistic knowledge makes the Chinese language processing model more interpretative.

With the development of intelligent information technology, automatic generation of questions and automatic verification of answers have become one of the main functions of the intelligent tutoring systems. Although some existing approaches to automatic generation of questions and automatic verification of answers are introduced in the literature, these approaches only allow to generate very simple objective questions and verify user answers with very simple semantic structure. So, this article proposes an approach for designing a general subsystem of automatic generation of questions and automatic verification of answers in intelligent tutoring systems built using OSTIS technology. The designed subsystem allows to automatically generate various types of questions based on information from the knowledge bases and multiple question generation strategies, and the subsystem can also automatically verify the correctness and completeness of user answers in the form of semantic graphs. Compared with existing approaches, the subsystem designed using the approach proposed in this article can not only generate various complex types of questions, such as multiple-choice questions, fill in the blank questions, questions of definition interpretation, etc., but also verify user answers with complex semantic structures.

The paper is presenting an updated edition of the second version of Theory for Automatic Generation of Knowledge Architecture (TAPAZ-2) and proposes a new approach to solving the problem of automatically identifying the semantic equivalence of text documents and borrowing scientific ideas in order to curb the spread of plagiarism and prevent clogging the information space under the conditions of its globalization.

We consider the use of ontological knowledge from user profile (domain of interests, current tasks, experience and competences etc.) for semantic retrieval. Domain ontologies, Wiki resources and task thesauri used in general technological chain of user-oriented semantic retrieval can be generated independently by different applications and are integrated with the help of the Semantic Web standards. Open information environment is considered as an external data base with great volumes of heterogeneous and semi-structured information that can be transformed into ontologies. Semantic Wiki resources provide generation of ontology for selected set of Wiki pages that formalizes their knowledge structure and explicitly represents its main features. Semantic similarity evaluations and knowledge about typical information objects of resources are used for selection of Wiki pages pertinent to user task. Such Wiki-ontology elements as classes, property values of class instances and relations between them are used as parameters for the quantitative assessment of semantic similarity. Task thesaurus that represents current task is generated on base of domain ontology and task description. The semantic retrieval system based on ontological representation of user needs is described. The set of domain concepts that are semantically similar to currents Wiki page can be used as a base for task thesaurus.

Considered are the questions of designing a system for modeling the reasoning of a cognitive agent, capable of making conclusions based on its knowledge and observations of the external environment, solving problems in a hard enough real-time mode. To work in this mode, the existence of a critical time threshold is established, which is set to solve the problem facing the agent. Exceeding the threshold is fraught with grave, sometimes catastrophic consequences and for the agent is unacceptable. The formal basis of the modeling system (cognitive process control) is a logical system - extended step theory, that combines the concepts of active temporal logic and logical programming. Among the original methods proposed by the authors in the work, an approach to combining the concepts of active logic and logical programming in one logical system should be noted; an approach to constructing a consistent declarative semantics for extended step theory of active logic; a method of formalizing temporal, nonmonotonic reasoning of an agent using extended step theory of active temporal logic; a method of granulating time in a logical system to formalize meta-reasoning. A subclass of temporal logic is considered, oriented to application in real-time systems. Additionally, the issues of managing the agent’s cognitive process in hard real-time, eliminating anomalies (unforeseen situations), and applying the temporal logic of branching time are investigated in more detail.

The interactive computer system GeoBazaDannych is a complex of intelligent computer subsystems, mathematical, algorithmic and software tools for filling, maintaining and visualizing input and output data, creating continuously updated computer models. The purpose and functionality of the main components of the system GeoBazaDannych are briefly described. Examples of interactive formation of digital models of geological objects in computational experiments that meet the intuitive requirements of the expert are discussed. Methodological and algorithmic solutions, corresponding special tools of the system GeoBazaDannych for the formation of digital distributions that meet the requirements set by the expert are noted. The results of comparison with standard solutions in the complex “Generator of the geological model of the deposit” are presented. Examples of approximation and reconstruction of the digital field and its interactive adaptation by means of the system GeoBazaDannych are given and discussed; the obtained solutions and their accuracy are illustrated by maps of isolines.

The paper provides approach for building neural network data processing technology based on deep belief networks. A neural network architecture, focused on parallel data processing and an original training algorithm implementing the annealing method, is proposed. The approach effectiveness is demonstrated by solving the image compression problem as an example.

The problem of restoration of the power grid after shutdowns requires extensive heterogeneous knowledge, has high combinatorial complexity, many limitations and conditions, including limitation on the decision-making time. Under such conditions, traditional abstract-mathematical methods are irrelevant to the complexity of the control object, and solving the problem by expert team is irrelevant to its dynamism. In this regard, hybrid intelligent multi-agent system that model collective heterogeneous thinking processes during the decision-making under the guidance of a facilitator are proposed to solve problems in dynamic environments, in particular distribution grid restoration-planning problem. The paper discusses the model, the functional structure, and the collective heterogeneous thinking protocol of such systems.