Название: Knowledge Integration Methods for Probabilistic Knowledge-based Systems
Автор: Van Tham Nguyen, Ngoc Thanh Nguyen, Trong Hieu Tran
Издательство: CRC Press
Год: 2023
Страниц: 203
Язык: английский
Формат: pdf (true)
Размер: 22.5 MB

Knowledge-based systems and solving knowledge integrating problems have seen a great surge of research activity in recent years. Knowledge Integration Methods provides a wide snapshot of building knowledge-based systems, inconsistency measures, methods for handling consistency, and methods for integrating knowledge bases. The book also provides the mathematical background to solving problems of restoring consistency and integrating probabilistic knowledge bases in the integrating process. The research results presented in the book can be applied in decision support systems, semantic web systems, multimedia information retrieval systems, medical imaging systems, cooperative information systems, and more. This text will be useful for Computer Science graduates and PhD students, in addition to researchers and readers working on knowledge management and ontology interpretation.

Today, Artificial Intelligence (AI) has been used in applications to solve specific problems in many fields such as knowledge-based systems (KBS), speech recognition, natural language processing (NLP), artificial vision, robots, neural networks with specific applications such as personalized shopping, AI-powered assistants, fraud prevention, administrative tasks, automated to aid educators, creating smart content, voice assistants, personalized learning, autonomous vehicles. In AI, each KBS is a computer system with the ability to think and make decision as a human expert. They are designed to deal with a large range of problems from Computer Science and engineering to social sciences such as economics, politics, and law. There are several well-known knowledge bases being developed and widely applied, namely DBPedia, Google’s Knowledge Graph, Neil, Open IE, Probase and Yago. When developing knowledge-based systems, several mathematical models such as Bayesian network models and Markov network models have been selected to represent a knowledge base. Recently, Islam et al. used applications of the fuzzy-Bayesian models to build a KBS for the cost overrun risk assessment of the power plant project. Blondet et al. also use Bayesian networks as an inference engine to build a special KBS.

The knowledge base is supported by a specific ontology. The inference engine reasons on the knowledge to assist designers in configuring numerical design of experimental processes efficiently. This system depends on an ontological model to integrate existing expert knowledge and to discover new knowledge from these analyses. The KBS also depends on the inference engine to combine knowledge fast. Enrique et al. proposed typical architecture of a KBS consisting of many of the basic components. Two types of problems that KBS could solve are deterministic and stochastic problems. Therefore, KBS can be classified into two main types: deterministic and stochastic KBS.

In this section, we make a survey of typical applications of uncertainty resolution and knowledge integration in AI including Machine Learning, Recommendation Systems, and Group Decision-Making. In Machine Learning research field, depend on the representation of knowledge, the applications of uncertainty resolution and knowledge integration is classified in three groups, namely Knowledge Graph, Probabilistic Relations, and Logic Rules. In the group of applications in Knowledge Graph, there are two typical works as follows: In " Knowledge-powered Deep Learning for word embedding", by arguing that text usually contains incomplete and ambiguous information, it causes the difficulties for applying Deep Learning to solve natural language processing (NLP) tasks, Bian et al. studies the capacity of leveraging morphological, syntactic, and semantic knowledge is performed to achieve high-quality word embeddings. This knowledge-powered Deep Learning is proven to enhance the effectiveness of word embedding. Multi-view Factorization AutoEncoder, a framework with network constraints to integrate multi-omic data with domain knowledge is studied by Ma et al.