Military event detection is one of the main tasks of military information extraction, which aims to identify event information in unstructured military text. Event detection is the first step of military event detection. Traditional military event detection methods are difficult to solve the problems of insufficient artificial features, inaccurate Chinese word segmentation in military field, and insufficient utilization of entity relationship features between sentences. Therefore, based on the pre-trained model (BERT), this paper proposes a military event detection method which combines BiGRU and attention mechanism. The language model is trained to construct a vector representation method combining word vector and position vector. BiGRU neural network is used to learn the above semantic features, and the attention mechanism is integrated to improve the expression ability of semantic features of sentences. This paper constructs a certain number of military text corpus. The experimental results show that our model improves the efficiency of military event detection compared with the traditional non attention model.