01Introduction to Knowledge Graph
The knowledge graph is a large-scale semantic network, a semantic representation of the real world. Entities are represented as nodes, and the attributes of entities and the relationships between entities are represented as edges, forming a web-like graph structure. This structured form is recognizable by humans and machine-friendly, facilitating machine understanding. The large-scale concepts, attributes, and relationships between entities in the graph provide rich semantic information and connections, inherently possessing various features of graphs, allowing for relevant operations and applications. Once constructed, it can also serve as background knowledge directly for downstream applications. Initially proposed by Google, the knowledge graph was primarily used to enhance the capabilities of search engines and improve search quality. Due to its unique characteristics, the knowledge graph plays an important role in various aspects of artificial intelligence.
02Building Subject Knowledge Graph in EducationKnowledge graph construction is typically application-oriented. Before constructing the graph, it is necessary to clarify which problems need to be solved, what knowledge is required to address these problems, and then design the ontology of the knowledge graph to cover this knowledge. First, let’s introduce some common application scenarios and logic of the educational knowledge graph, design the ontology of the educational graph based on these applications, and then further construct the knowledge graph. 1. Application Logic of Educational Knowledge Graph The educational knowledge graph is centered around subject knowledge, establishing hierarchical relationships among knowledge points of various subjects, the associations between knowledge points, and the sequential relationships among different knowledge points, forming a subject knowledge graph. Using this graph, the relationships between knowledge points can be visually displayed to students, making it clear and intuitive, naturally helping students build a knowledge system, review key knowledge points, discover associations between knowledge points, and assist students in summarizing and consolidating knowledge, eliminating blind spots. Once constructed, the subject knowledge graph can be associated with teaching resources (textbooks, test questions, handouts, teaching videos, exam papers, etc.), and through user information and learning records, establish connections between knowledge points and users. The knowledge graph can more accurately depict students’ knowledge mastery and resources, enabling precise analysis of user learning situations, learning path planning, and personalized recommendations for learning resources. It can also help teachers better understand students’ learning situations, optimize teaching methods, and adjust teaching strategies. By associating with teaching research materials, it can actively recommend teaching research to enhance the efficiency and quality of lesson preparation for teachers. A teaching assistance Q&A system centered on knowledge graph Q&A technology can effectively reduce the burden of simple repetitive questions on teachers and significantly meet students’ Q&A needs.2. Educational Knowledge Resource Construction Centered on Subject Knowledge Graph The construction of educational knowledge resources centered on the knowledge graph establishes connections between domain knowledge, knowledge points, and various educational resources such as different versions of textbooks, teaching aids, handouts, videos, and test questions, forming a cohesive network. Using these connection networks supports upper-level applications. 
In the subject knowledge graph of the education field, the relationships between knowledge mainly include: hierarchical relationships, primarily between parent and child concepts, and between concepts and entities. The concept graph expresses the relationships between concepts in the education field. The conceptual content in the education field is often abundant, and these relationships form the entire knowledge framework. Containment relationships, where several specific small test points fall under a knowledge point, represent the overall and part relationship. Sequential relationships can be used for learning planning. In different subjects, there are also some special relationships, such as mutual exclusion and causality, which need to be refined by domain experts and knowledge engineers during the construction of graph resources. Knowledge in the educational graph also has rich attributes, such as common attributes like “test points,” “difficult points,” “easy-to-make mistakes,” and “exam syllabus requirements.” Different subjects have specific fine-grained attributes, such as “definition,” “properties,” “area formula,” and “perimeter formula” in mathematics.03Application of Knowledge Graph in Smart Education1. Precision User Profiling Based on Educational Knowledge Graph Precision user profiling based on the educational graph extracts key information by analyzing user behavior information and its connections with various resource objects to portray users. User profiles based on knowledge graphs can enhance user profile data, making user characterization more comprehensive and accurate.1. Precision User Profile User profiling technology is already quite familiar to everyone, aimed at better characterizing users for understanding purposes, fundamentally involving “tagging.” The accuracy of user profiling determines whether the understanding of users is correct. In practice, there are mainly two issues to resolve in user profiling: first, the incompleteness of profile data, and second, the inaccuracy of profile data. To address these issues, knowledge graphs can be utilized. The nodes in the knowledge graph abstract educational domain knowledge, covering a sufficient number of entities and concepts, serving as a source for user profile tags. Its high quality ensures more accurate tagging. These tags have connections, and the graph contains rich semantic relationships, helping machines understand the meanings of these tags. The friendly structure facilitates better human understanding and the intuitive discovery of relationships among tags. Algorithms such as tag propagation and cross-domain recommendations can be used to mine more precise tags to describe users, enriching user tags and enhancing user accuracy. Additionally, when creating user profiles, user profiles can be presented as a product visualization, generating personalized, dynamically changing user knowledge graphs based on the relationships in the graph.2. Precision Learning Situation Analysis The knowledge graph enables more precise learning situation analysis. Traditional educational experts (teachers’ experience) rely on experience to assess learners’ knowledge and ability states, lacking the integration of educational measurement concepts, resulting in significant randomness and subjectivity. Based on knowledge graphs and big data analysis methods, objective learning processes of learners can be mined, analyzing from multiple dimensions. The data can be mined from numerous dimensions, not limited to test scores, error books, and learning records, revealing knowledge mastery and weak knowledge as explicit features, while also uncovering deeper learning speeds, learning preferences, cognitive levels, and other implicit features. This makes the analysis results more personalized and objective. For unmet learning goals, knowledge graphs can perform root cause analysis, identifying weak points and related knowledge points, effectively filling gaps. The diagnostic process has better adaptability and personalization.2. Improving Teaching Quality and Efficiency The knowledge graph assists in teaching applications, helping teachers with lesson preparation, teaching research, question creation, and test analysis. The system can recommend similar relevant materials (lesson plans, course explanation planning, assignments, etc.) to teachers to improve their teaching efficiency. The search based on the graph can also return the required content more accurately.1. Intelligent Lesson Preparation Utilizing the subject knowledge graph, the subject knowledge points are associated with textbooks, handouts, exercises, etc., and based on the teacher’s teaching progress and textbook version, continuously push lesson preparation resources that meet teaching needs, enabling quick and accurate access to the resources teachers require, improving lesson preparation efficiency and quality. In addition to lesson preparation, the subject knowledge graph can also be used as background knowledge to assist in intelligent question generation and exam analysis. This greatly enhances the efficiency of teaching research.
2. Smart Classroom Through precise analysis of learning situations, the system recommends relevant reinforcement exercises, formulates targeted teaching strategies, and enhances teaching relevance, conducting precise teaching. Before, during, and after class, data mining and intelligent capabilities are comprehensively utilized. Before class, using data mining technology to obtain student learning situation data, teaching strategies are formulated to make decisions data-driven. During class, targeted teaching is conducted based on the formulated teaching strategies, explaining knowledge points and facilitating group discussions. Utilizing the knowledge graph to visualize the inherent connections of knowledge helps students build a deeper understanding of knowledge. After class, relevant exercises are recommended based on students’ learning situation, learning abilities, and personalized targeted practice exercises, reinforcing mistakes. This enhances the quality and relevance of classroom teaching. Dynamic data analysis and dynamic learning situation diagnostics run throughout the entire teaching process, achieving tailored instruction and making teaching decisions data-driven and intelligent.3. Deep Reading Deep reading based on knowledge graphs aims to achieve intelligent and comprehensive understanding of the relationships between knowledge. Using entity linking technology to identify and connect entities in electronic publications can display current knowledge information in the form of knowledge cards. It can also link to other related knowledge and recommend related knowledge, helping users connect knowledge. This greatly promotes users’ comprehensive understanding of knowledge. Deep reading can be applied not only in the education field but also effectively in knowledge management and the publishing industry.The core technology relied upon for deep reading is entity linking technology. Currently, our knowledge workshop’s entity understanding service allows machines to understand entities in text, making entity search and semantic search possible, achieving over 90% accuracy and recall in general domains.4. Q&A RobotsEducational robots have become an important application in the education field. Utilizing education robots centered on Q&A systems can achieve course Q&A, knowledge retrieval recommendations, teaching management, and a series of teaching tasks. This not only relieves teachers of burdens and pressures but also solves students’ practical problems. An excellent comprehensive teaching robot integrates task-oriented Q&A, chatbots, knowledge-based Q&A, search recommendation systems, and other system modules, while also possessing multi-turn Q&A capabilities. Knowledge graphs play an important role in understanding query questions, generating language guided by knowledge, and are core to KBQA. 
04Evolution Path of Intelligent Education Centered on Knowledge Graph The knowledge graph increasingly takes on the mission of assisting industry intelligence. Exploring the evolution path of industry intelligence based on knowledge graphs is thus crucial. After years of practice, this path has become clearer, presenting an iterative development model of knowledge resource construction and knowledge application. The basic principles for implementation are: overall planning, application-led, and using applications to promote construction. The iterative development path of knowledge resource construction and knowledge application. 
1. Application-led, Using Applications to Promote Construction The implementation of intelligence in the education industry must follow the implementation plan of the general industry, being application-oriented. There are numerous educational application scenarios, requiring overall planning and gradual implementation. Based on the current business development needs and data, technical foundations, suitable application scenarios should be selected, and capabilities for constructing graphs should be built in specific applications, directly generating business value. It is unwise to pursue graphs merely for the sake of graphs without concrete business support; unless it is a leading company’s AI Lab not seeking short-term profits, it is difficult to gain adequate internal support. Only by generating concrete business value can there be sufficient motivation to promote the long-term and steady development of intelligent transformation projects centered on graphs.2. Iterative Development In the implementation of knowledge graphs in the education industry (and similarly in vertical industries), compared to knowledge application, the acquisition of knowledge and the construction of knowledge resources are greater bottlenecks. Knowledge resource construction is a long-term task that cannot be accomplished overnight; it requires solid progress in building knowledge resources and accumulating knowledge bases, without shortcuts. An iterative spiral development model is adopted, starting from specific points to broader areas. In each iteration cycle, the appropriate principle must be grasped, prioritizing application scenarios with better expected outcomes, reasonably controlling the boundaries and volume of knowledge, constructing knowledge resources centered on knowledge graphs, and carrying out corresponding knowledge applications. Feedback from internal and external users should be used to improve corresponding applications and knowledge resource construction. Once specific applications show initial results, the scope can gradually expand to more application scenarios, constructing more knowledge resources. The entire process continues to iterate until comprehensive intelligence is achieved. Cognitive intelligence is the key to advanced artificial intelligence, and the realization of cognitive intelligence relies on knowledge graphs. Today, there are increasing applications of vertical industry intelligence based on knowledge graphs. Many application scenarios in the education field have shown good results. However, the overall application of cognitive intelligence in the education field is still in the development stage, with many areas for improvement. For instance, in graph construction, the demand for multimodal knowledge graphs in educational knowledge graphs is evident. The complexity of knowledge in the education field increases the difficulty of segmenting knowledge point granularity, and the richness of relationships between knowledge points also requires continuous optimization. Overall, the application of knowledge graphs in educational intelligence has a tremendous impact and a bright future. With the further maturity of technology, support from national policies, continuous capital investment, and ongoing innovation by commercial companies, educational intelligence is expected to achieve remarkable results in the coming years.
Source: Hexi University
The above images and text are valuable for sharing, and copyright belongs to the original author and source. The content reflects the author’s views and does not represent this public account’s agreement with these views or its responsibility for their authenticity. If there are copyright issues, please contact us in a timely manner.