Multi-task Chinese entity naming recognition method

Abstract

The invention discloses a multi-task Chinese entity naming recognition method, which comprises the following steps of: (1) firstly, preprocessing data, dividing a data set and setting a labeling task; (2) carrying out feature extraction on the input main task data and auxiliary task data through BERT; (3) for the main task and the auxiliary task, performing classification training on word vectors by adopting a double-layer LSTM neural network model comprising input, hiding and output; (4) performing full connection on the trained hidden layer information of the auxiliary task and the main task through an attention mechanism layer; (5) finally, considering global label information in the sequence through a CRF layer, and outputting an optimal label sequence; and (6) performing performance evaluation on the trained model through the verification set. According to the method, researchers can be helped to efficiently obtain valuable information and knowledge from massive Chinese text data, the problem that time and labor are consumed in manual information extraction is effectively solved, and the method has important significance in further text mining work.

Description

technical field [0001] The invention relates to the technical field of text mining, in particular to a multi-task Chinese entity naming recognition method. Background technique [0002] Nowadays, the problem of entity naming recognition in English text has been widely studied. However, Chinese NER still faces challenges such as Chinese word segmentation, and it is often difficult to define the composition of a word. In the past Chinese NER tasks, the recurrent neural network (RNN) was often used to improve the performance of the model on entity classification tasks, but RNN has the problem of gradient disappearance and gradient explosion in the long-distance training process. The long-term short-term memory model (LSTM) can be used in Longer sequences have better performance, and a simple tuning trick for LSTM units in RNNs can significantly reduce overfitting. The neural network model combining Bidirectional Long Short-Term Memory (BiLSTM) and Conditional Random Field (CR...

AI Technical Summary

Problems solved by technology

[0003] Because the entity boundaries of Chinese named entity recognition are difficult to divide, there are problems such as word segmentation errors, out-of-vocabulary (OOV), etc., BILSTM is not ideal as a feature extractor.

The pre-training of word vectors is based on the feature extraction of words and characters, ignoring the relevant information of words in the context, and extracting a static word vector that does not contain context, which also leads to the model’s ability to recognize entities. Decline

Images