Named entity identification method based on adversarial training

Abstract

The invention discloses a named entity identification method based on adversarial training. The method comprises the following steps: obtaining relevance features among judicial field characters through RoBERTa model training and Bi-LSTM training; splicing the two relevance features together, and predicting a training sample by using a conditional random field model to obtain a prediction result;according to the method, external word vectors and word vectors of different dimensions can be introduced to be combined with judicial domain text word mixing vectors of different dimensions, confrontation disturbance is carried out on the mixing word vectors in a judicial domain text, and the accuracy of model recognition is improved.

Description

technical field [0001] The invention belongs to the technical field of named entity recognition, and in particular relates to a named entity recognition method based on confrontation training. Background technique [0002] Named entity recognition has been widely used in various fields, and different fields have been optimized for named entity recognition. In traditional named entity recognition, it takes a lot of people to extract features for specific fields, and use probability Graph model for named entity recognition. With the rise of deep learning in recent years, various fields have used deep learning methods to conduct a lot of exploration on named entity recognition. At present, there have been a lot of exploration and practice in the fields of finance, medical treatment and law. , reducing a lot of labor costs and improving the accuracy rate. How to use this information is particularly critical. In the use of named entity recognition technology, entities with specif...

AI Technical Summary

Problems solved by technology

[0007] The existing domain-specific named entity model uses the combination of Bi-LSTM and CRF model, but the ability of the model to extract features is not strong enough, and the modeling in Bi-LSTM is simply from left to right, or from right to Modeling on the left and splicing the hidden states together, but the disadvantage of this is that it can only use the information of the above or the following, and cannot use the information of the above and the following at the same time

Moreover, the number and number of texts in a specific field are limited, and there is no large amount of data to improve the performance of the model

[0008] With the emergence of the BERT model, it has been gradually applied in various fields, but it has not been applied in specific fields. However, the words brought by BERT and the subsequent model RoBERTa are independent of each other, and fine-tuning will bring the model Disadvantages such as loss of performance, the scale of data is large, and the accuracy of the model cannot be improved

Images