[Objective] In the era of big data and artificial intelligence, this study utilizes domain knowledge graph-based intelligent technology to achieve precise natural language semantic parsing for human-machine interactive intelligent question-answering services.

[Methods] In this study, a pipeline approach is employed to construct a knowledge graph-based retrieval-based question-answering system. Firstly, the Roberta pre-training model and data augmentation techniques are combined with the knowledge graph to address the issues of low accuracy in question classification and named entity recognition in low-resource environments. Furthermore, based on the characteristics of military entities, an entity linking technique using three-dimensional features is proposed. To solve the problem of relation matching between simple and some complex intent questions, the Roberta pre-training model and dependency parsing are utilized. Finally, answer extraction is accomplished through the application of heuristic rules.

[Results] The question-answering method proposed in this study achieved an average accuracy of 91.94% in the evaluation, indicating the practicality and accuracy of the system in providing efficient military question-answering intelligent services. This question-answering method successfully fulfills the requirements for intelligent services in the military domain.

[Limitations] Due to the limited knowledge scale of the existing military knowledge graph, the supported scope of question-answering needs further expansion.

[Conclusions] This study provides an efficient and precise human-machine interactive military intelligent question-answering service supported by intelligent technology.