Research on Multimodal Knowledge Graph Construction Method Based on Conditional Random Field Model

• Published in: 2024 7th International Conference on Education, Network and Information Technology (ICENIT) pp. 60 - 67
• Main Authors: Peng, Kanghua, Qiao, Yifang, Shi, Jincheng
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.08.2024
• Subjects: Analytical models; Conditional Random Field (CRF) Model; Conditional random fields; Data mining; Dependency Syntactic Analysis; Education; Entity Relationship Extraction Algorithm; Knowledge engineering; Knowledge Graph; Knowledge graphs; Python; Search problems; Software; Syntactics
• DOI: 10.1109/ICENIT61951.2024.00019

In an effort to enhance the efficiency and decision-making levels of higher education teaching management, and to address the fragmented, disorganized, and scattered state of online and offline educational resources, a new approach to constructing educational resources in emerging fields has been explored. The study proposes a method for entity extraction suitable for the software technology specialty domain, designs a Conditional Random Field (CRF) model, and employs new word discovery plus named entity recognition, along with dependency syntax analysis and entity relationship extraction algorithms for knowledge fusion. This leads to the visualization of knowledge graphs, building semantic associations within the entire discipline's knowledge points, and constructing a multimodal educational knowledge graph in the field of database technology. Using the software technology specialty group's platform course on database technology as a carrier, the method is implemented across three scenarios: offline textbooks, online MOOCs, and IT blogs. Among these, text recognition has achieved 100% accuracy, recall probability, and F-value on some pages, successfully completing the construction of subject and course knowledge graphs. The results show that the knowledge graph can visualize the relationships between course knowledge units, optimize course settings, meet the needs of intelligent search for teaching resources and personalized recommendation of course resources, solve the problem of resource information disorientation, make the use of educational resources more efficient, and achieve the construction goals of supporting teaching resources.