Abstract: 【Purpose/significance】Realize the automatic learning and identification of domain ontology classification relations, and
solve the problem of the automatic construction of domain ontology knowledge framework structure.【Method/process】Through the anal? ysis and summary of the research status and existing problems of automatic recognition of domain ontology classification relationship at home and abroad, the paper studies and builds the domain ontology classification relationship automatic recognition model based on the open source and advanced deep learning pre-trained models BERT. Then, the paper takes the field of resources and environment as an example to conduct experimental research and evaluation analysis.【Result/conclusion】The classification model based on BERT can realize the automatic identification of domain ontology classification relationships. The identification methods and processes are extremely versatile and portable, and the identification accuracy is greatly improved compared to traditional methods.【Innovation/limi? tation】The paper fines tuned and generalized BERT，improves the generality and accuracy of domain ontology classification relation recognition model. However, Due to the quality constraints of the classified annotation corpus, the model accuracy has not yet reached its peak and needs to be further optimized and improved.