Online collaborative learning can effectively improve knowledge construction level, critical thinking and communication skills of students which has become one of the important learning methods for students. A large number of interactive texts will be generated in the process of online collaborative learning, while teachers are facing with the problem that it is difficult to know the learning situation of individual students, learning groups even whole class in real time, so it is difficult to provide learning support and feedback. The research aims to automatically construct domain knowledge graphs in real time based on interactive texts, so that teachers can understand the learning situation of individual students, groups and class. The research content is divided into two parts: (1) research on domain knowledge graph construction technology; (2) the case study of domain knowledge graph application.

(1) The research on construction technology of domain knowledge graph. By designing an online collaborative learning activity based on the development of a three-person version of the Snake C language game, using VSCode as the online collaborative learning platform, we recruited 90 students were divided into 30 groups for online collaborative learning activities. Based on the 14193 interactive texts generated in the online collaborative learning process of 90 students, two entity recognition methods are used to identify the knowledge entity, then we construct relationship recognition. Finally, there are 7 domain knowledge graphs includes the initial knowledge graph, the activated knowledge graph and unactivated knowledge graph of student, learning group and class are generated in the Neo4j graph database. In the first entity recognition method, an entity recognition model that combines three models of text classification, entity recognition and keyword matching is used, and the overall accuracy is 78.56%. In the second method, using a synthetic model of sequence recognition, text classification and keyword matching, the overall accuracy of the model is 87.27%. The second method is significantly better than the first in terms of accuracy of entity recognition and the rationality of model design. Therefore, the second method is selected as the final domain knowledge graph construction model of this research.

(2) The case study of domain knowledge graph application. Recruiting 6 people into 2 groups to conduct online collaborative learning activities on the VSCode platform. Based on the domain knowledge graph construction model in the first research, 7 knowledge graphs are automatically constructed for students in real time. We tested the accuracy of the domain knowledge graph construction model in the real learning situation, and learned learners' attitudes and feelings towards the 7 knowledge graphs. Through the focus interview, it can be learned that the 7 automatically generated knowledge graphs have a high accuracy, and their effect on students' basic knowledge learning is greater than their programming ability improving. The 7 automatically generated knowledge graphs are usable, easy to use and friendly, they will not bring high cognitive load to students. At the same time, the students who participated in the experiment agreed that the in-depth processing of the 7 knowledge graphs, with reasonable and effective teaching activities and teaching time, will help them play a greater role in learning assessment and learning support.

This research is the first innovative one in the field of online collaborative learning to automatically construct a knowledge graph. It proposes a domain knowledge graph construction method that conforms to the characteristics of online collaborative learning interactive texts, which has strong academic value and practical application value.