Abstract
AbstractNamed entity recognition is a basic task in the field of natural language processing. The existing named entity recognition model has the problem of poor recognition effect on specific domain data sets. Aiming at the inconsistency of sequence labeling caused by the shortcomings of the traditional BIO (B-begin, I-inside, O-outside) annotation method, a named entity recognition model based on transformer is proposed, which uses the feature extraction of encoder in transformer and the labeling method of limited interval. Firstly, the BERT pre-training model is used to perform word-level dynamic vector representation of the input text. Then the vector representation of the text is intercepted by the span-based labeling method, and the feature of text and candidate named entity is extracted by self-attention mechanism. The results of experiment on the dataset in the industrial informatization field show the effectiveness of the model.KeywordsNamed entity recognitionTransformerSelf-attention
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call