Improved Chinese named entity identification method based on Lattice-LSTM

Abstract

The invention relates to an improved Chinese named entity recognition method based on Latte-LSTM, and belongs to the technical field of language processing. The method comprises the following steps: S1, model constructing; s2, feature input; s3, feature extraction; s4, label prediction; s5: result evaluation. According to the invention, an improved LSTM structure is adopted; hidden information ofthe sentence in complete semanteme is calculated; meanwhile, the consideration of global information of the whole sentence is also added; more importantly, the sentence structure is opened from the perspective of sentence structure; the defect that an LSTM structure only pays attention to hidden information of text meaning and does not consider sentence structure information is overcome, and afterthe Transformer structure is fused, the model can understand logicality behind complex sentences to a certain extent, so that named entity categories in the sentences are helped to be recognized.

Description

technical field [0001] The invention belongs to the technical field of language processing, and relates to an improved Chinese named entity recognition method based on Lattice-LSTM. Background technique [0002] Named entity recognition was first organized by Grishman and Sundheim at the Sixth Information Understanding Conference in 1996. The development of named entity recognition can be roughly divided into three stages, which are the early primary stage of entity class recognition using manual rules, and the later stage. Around 2000, the machine learning method combined with the probability model was used to identify the advanced stage of the entity class, and then to the currently popular in-depth stage based on the deep learning method combined with the language model, each stage has some points worth learning, as follows Introductions are made from these three stages. [0003] The early method of using manual rules was combined with dictionary construction and standar...

AI Technical Summary

Problems solved by technology

[0013] In view of this, the purpose of the present invention is to provide an improved Chinese named entity recognition method based on Lattice-LSTM, to solve the problem of the accuracy of the recognition effect of Chinese named entity recognition, while utilizing the advantages of LSTM to process sequence sequences, and the Transformer structure can Deal with the advantages of longer distance sentences and jointly calculate the probability of named entities, so as to achieve the purpose of improving the recognition effect

Images