Abstract: In order to achieve automated construction of the knowledge graph of sheep disease, it is necessary to extract sheep disease triplet information from a large number of sheep disease text corpora. Aimed at the low efficiency and accuracy of the current pipeline-based entity relationship extraction model, a sheep disease entity relationship joint extraction model based on R-BIO annotation is proposed based on a self-built sheep disease dataset. This model utilized BERT encoding and concatenation mechanism, while extracting entities and relationships from text, avoiding the inefficiency and error propagation issues of pipeline methods. The experimental results show that the model outperforms the pipeline method in sheep disease triplet extraction tasks, with accuracy and recall rates of 74.62% and 73.67%, respectively.