Relationship classification method based on combination of pre-training model and syntactic subtree

The invention discloses a relationship classification method based on a pre-training model in combination with syntactic subtrees. The method comprises the following steps: firstly, constructing a word vector, a sentence representation vector and an entity vector by utilizing a BERT pre-training model; in order to combine syntactic information, a Spacy toolkit is used for carrying out dependency syntactic analysis on sentences, then an analysis result is preprocessed, and edges and edge categories are obtained. When syntactic information is combined, a recurrent neural network RvNN is used for recurrent calculation, a representation vector of each sub-tree is obtained, and the purpose of the step is to obtain topology information, semantic information and edge category information of the syntactic dependent tree. And performing maximum pooling on the representation vector of each sub-tree to obtain the representation vector of the tree. The entity vector, the sentence representation vector and the representat