Article Preview
TopIntroduction
At present, the knowledge data obtained by most workers in the ceramic domain through the above methods are not only time-consuming and labor-intensive, but also may have the problem of knowledge loss. In addition, the acquired knowledge also lack of unified expression and effective organization, which is difficult to be directly used in machine learning of data in the ceramic field. On the other hand, in the field of ceramic data, because the data are too scattered, there is a small sample problem, which also brings challenges to machine learning.
How to collect, organize, and manage this scattered knowledge in the field of ceramics and how to support the scientific research of ceramics are urgent problems to be solved.
It is well-known that, in 2012, Google proposed a new search function—knowledge graph-based search—aiming to improve search performance; the concept of knowledge graph has gradually become known to everyone. The knowledge graph explicitly precipitates and correlates domain knowledge, which can well solve the characteristics of scattered, complex, and isolated data in the domain (Xu et al., 2016). The question-answering system deals with natural language questions. The first thing to do is to identify the named entity of the question, and then retrieve the knowledge from the knowledge base and extract the corresponding answer. Knowledge graphs are used for knowledge retrieval and answer extraction due to their knowledge reasoning capabilities, which provide Q&A systems with a knowledge base that has a recommendation mechanism and higher accuracy. The two major links provide a knowledge base with a recommendation mechanism and higher precision for the question-answering system (Xia, 2016).
In recent years, knowledge graph technology has been applied in many fields and achieved remarkable results (Wu et al., 2021), such as smart finance, smart medical care, smart education, and smart e-commerce (Fu et al., 2021; Hang et al., 2019). However, in the ceramic field, the related research of knowledge atlas is still in the initial stage (Cao, 2021; Katiyar & Cardie, 2017). Besides, only a few researchers have carried out exploration on ceramic knowledge extraction (Li, 2020; Zheng, Hao et al., 2017; Zheng, Wang et al.,2017), carried out the extraction research of general type knowledge such as ceramic entity and relationship, rarely involved in the acquisition method of knowledge with ceramic field characteristics such as process flow, and almost no research on establishing ceramic knowledge atlas and applying it to ceramic machine learning.
The application of knowledge graph technology in the ceramic domain can effectively utilize the advantages of knowledge graphs in domain knowledge learning, organization, and reasoning, which is helpful to solve the challenging problems of machine learning in the ceramic domain.
In the process of knowledge graph construction, entity and relationship extraction is the foundation, entity alignment is the guarantee of knowledge fusion, and knowledge graph completion is the core to improve the quality of knowledge graph. In the process of entity and relationship extraction, semantic analysis and understanding of content information cannot be separated from the support of a semantic knowledge base (Li & Ji, 2014; Wang et al., 2016), and the machine can understand human language through a semantic knowledge base, thus becoming more intelligent. However, most of the traditional semantic knowledge bases are oriented to general fields, which cannot meet the knowledge requirements of natural language processing systems in specific fields (Guo et al., 2018; Zeng et al., 2018). Therefore, the authors will solve the related problems of ceramics by constructing knowledge graphs of the ceramic domain.