Chinese named entity recognition method based on BERT

The invention relates to the technical field of character coding, in particular to a BERT-based Chinese named entity recognition method, which comprises the following steps of: S1, coding a text sequence by taking a single word as an object to obtain a vector sequence XA; s2, according to words obtained through word segmentation and character information in the words, encoding and averaging are carried out, and a vector XB is obtained; s3, performing transformation fusion on the vector XA and the vector XB to obtain a vector XC; the Chinese named entity recognition method based on BERT has the beneficial effects that when the text is coded, not only is a single character coded, but also the relationship among the characters in the words is considered. Firstly, word segmentation is carried out on a character string sequence, then single characters in words obtained through word segmentation are coded and averaged to obtain an average vector, the single character vector and the average vector are subjected to t