Nested entity detection method based on pyramid hierarchical network

The invention discloses a nested entity detection method based on a pyramid hierarchical network, and belongs to the technical field of natural language processing. The nested entity detection methodcomprises the following steps: (1) encoding a word to obtain a word vector, and inputting the word vector into a first decoding layer as a representation of a text region only containing a single word; (2) the current decoding layer performs named entity recognition on the representation of each input text region, combines the representations of two adjacent text regions into a representation of anew text region, inputs the representation of the new text region into the next decoding layer, and repeats the step until the number of times reaches a preset layer number or the length of an inputtext; (3) training by using a proper optimization method corresponding to the pyramid hierarchical network; and (4) inputting a to-be-detected text into the trained pyramid hierarchical network to obtain a named entity recognition result. According to the nested entity detection method based on the pyramid hierarchical network, the problem that entity nesting exists in named entity recognition canbe solved.