Chinese knowledge graph construction method based on CasRel model

The invention relates to a relation triple extraction method in a knowledge graph construction information extraction task, in particular to a Chinese knowledge graph construction method based on a CasRel model. Comprising the steps of obtaining text data and constructing a data set in a specific form; chinese text data are input into a BERT Chinese pre-training model to obtain word vectors, the word vectors are input into a head entity tagging device, and after an Attention mechanism is used for forming head entity relation tuples, the head entity relation tuples are input into a series of specific relation tail entity tagging devices. And finally, in combination with a tail entity result identified in a relation cascade decoder, completing relation triple extraction. According to the method, a brand new relation extraction method fusing an Attention mechanism relation mapping function is adopted, the situation that overlapped triads and nested triads cannot be completely extracted through traditional relati