基于混合专家大模型提取的“基因-疾病-药物”三元知识图谱的设计

his paper focuses on extracting "gene - disease - drug" triplet knowledge from biomedical literature based on knowledge graphs. The core tasks include mining "gene - disease" associations from PubMed abstracts, extracting "compound - disease" entities and relationships from PubMed literature, and identifying drug - drug interactions fromDrugBank database texts. By employing the Mixture of Experts (MoE) model and going through steps such as data preprocessing, model fine - tuning (e.g., LoRA for efficient parameter - efficient fine - tuning) and result evaluation, this project utilizes large language models to structurally extract key knowledge from biomedical literature, thereby supporting medicalresearch and knowledge discovery