Research Of Entity Named Recognition Based On Neural Network

The main task of named entity recognition is to identify meaningful entities such as person,location,and organization from the text,which is one of the basic tasks of natural language processing.Named entity recognition plays an important role in many areas such as knowledge graph construction and information retrieval.However,the traditional named entity recognition method relies on the artificially constructed feature template,which requires a lot of manpower and material resources,and has limitations and deficiencies in processing large-scale text data.The use of deep learning for named entity recognition can reduce the workload of feature extraction and solve the problem of data sparsity.In order to construct the multi-language knowledge graph of the ten ASEAN countries,the first task is to identify the multi-language corpus.Therefore,this paper proposes a neural network-based named entity recognition method.The specific research contents are as follows:(1)A Chinese named entity recognition model based on BiGRU-CRF is proposed.The model first uses word2 vec to convert Chinese characters into low-dimensional dense vectors.Secondly,the BiGRU network is used to extract the semantic features of the vector.Finally,the optimal tag sequence is predicted and output through the CRF layer.To further enhancing the recognition effect,two hidden layers are added between the BiGRU and CRF layers.In the aspect of data preprocessing,a data set partitioning algorithm is proposed,and the text is more scientifically and rationally divided.The model is compared with several hybrid models on the ASEAN dataset.The experimental results show that the model has better recognition performance in the named entity recognition task.(2)When dealing with cross-language named entity recognition tasks,the disorder problem of words often occurs.The BiGRU-CRF model can not meet the actual needs.Therefore,the Self-Attention mechanism and CNN and BiGRU-CRF model are combined to construct a Self-Attention-based BiGRU-CNN-CRF cross-language named entity recognition model.Firstly,the GAN is used to train the cross-language word vector.Secondly,the semantic features of the vector are obtained through CNN and BiGRU.Finally,the optimal tag sequence is predicted and output through the CRF layer.The addition of Self-Attention mechanism in BiGRU can alleviate the problem of word order confusion in different language conversions,highlight the influence of keywords on entity labeling,and so improve the recognition performance of named entities.The experimental results show that the F1 value of the model in the task of cross-language named entity recognition in Spanish,Dutch and German are 72.73%,71.56% and 58.03% respectively,which are the optimal results.