KGSG: Knowledge Guided Syntactic Graph Model for Drug-Drug Interaction Extraction

The explosive growth of biomedical literature has produced a large amount of information on drug-drug interactions (DDI). How to effectively extract DDI from biomedical literature is of great significance for constructing biomedical knowledge and discovering new biomedical knowledge. Drug entity names are mostly nouns in specific fields. Most of the existing models can’t make full use of the importance of drug entity information and syntax information for DDI extraction. In this paper, we propose a model that can reasonably use domain knowledge and syntactic information to extract DDI, which makes full use of domain knowledge to obtain an enhanced representation of entities and can learn sentence sequence information and long-distance grammatical relation. We conducted comparative experiments and ablation studies on the DDI extraction 2013 dataset. The experimental results show that our method can effectively integrate domain knowledge and syntactic information to improve the performance of DDI extraction compared with the existing methods.