[Objective] Realize the automatic identification of the hypernym-hyponym relations of domain concepts to solve the problem of automatic acquisition and establishment of semantic relations between domain concepts in the automatic construction of domain ontology.

[Methods] The traditional unsupervised pattern-based method and the current advanced supervised-based projection learning methods are combined organically to apply to the automatic identification of domain concepts, and experimental research is carried out.

[Results] The method can identify the hypernym set of the domain concept. The identification accuracy in the medical field is 0.88, in the general field is 0.83, and in benchmark dataset BLESS is 0.85.

[Limitations] Affected by syntactic ambiguity, the quality of the corpus, the model accuracy has not yet reached its peak, and there are cases of misidentification.

[Conclusions] The model can find hypernym with different meanings of the same concept, and it also has a good identification effect on low-frequency words and named entities. In the future, consideration should be given to optimizing the identification method by appropriately reducing the weight of high-frequency top-level words and improving the quality of supervised corpus.