Research on Named Entity Recognition Based on Prompt Learning

Named Entity Recognition is the identification of named entities from a sequence of text. The pre-training language model is widely used in natural language processing tasks, but the named entity recognition task is different from the pre-training model task, so the pre-training model cannot be fully used. This article proposes a named entity recognition method based on prompt learning. Different templates are inserted into the input, and the pre-trained language model is used to transform the named entity recognition problem into a mask prediction problem. The mask prediction ability of the model is utilized to predict different mask positions, compensating for the different effects of the named entity recognition task and the pre-trained model task. External knowledge is integrated into the category mapping. Expanding the range of mapping words to improve recognition rate. Named entity recognition experiments were conducted on two datasets, WNUT16 and WNUT17. The appropriate template and category mapping words led to an F1 score of 68.96% in the WNUT16 dataset, which increased by 9.46 percentage points. The precision and recall rates were 64.42% and 67.72%, respectively. In the WNUT17 dataset, the F1 score was 69.62%, increasing by 9.17 percentage points. The precision and recall rates were 73.39% and 71.51%, respectively.