Zhuang Named Entity Recognition Based On Deep Learning

In the era of 5G communication,extraction of key information from the massive data on the Internet has become a research hotspot in natural language processing technology.Named entity recognition technology is to identify proper nouns from unstructured text.It plays an important role in natural language processing,especially for downstream tasks such as relation extraction and information retrieval.Meanwhile,artificial intelligence technology has been promoted rapidly with the increasing computing power and the emergence and optimization of new algorithms.Among the algorithm,some deep learning models have achieved excellent performance and become mainstream algorithms in named entity recognition tasks.As the language of the largest ethnic group in China,Zhuang language carries the wisdom of the Zhuang people.However,the survey found that the application of intelligent information processing technology in Zhuang language lags behind that of English and Chinese.Therefore,the general trend of informatization development in Zhuang language is to use natural language processing technology.Particularly,Zhuang named entity recognition,as a key technology in the development of Zhuang language informatization,is of great research significance and application prospects.However,relevant research shows that current research is not sufficient on named entity recognition in Zhuang language.Therefore,this paper launches research on Zhuang language named entity recognition technology by combining the advantages of deep learning and machine learning.The main research work and contributions are as follows:1.Due to the lack of relevant annotation datasets in the named entity recognition in Zhuang language,this paper collects Zhuang language news and other text corpora from the journals《Guangxi Nationality News》and 《March 3》,and performs cleaning and labeling work on the text.According to the current development status of Zhuang language and the characteristics of named entity recognition task,an annotated dataset suitable for Zhuang language named entity recognition research is constructed.2.Based on the Zhuang named entity recognition task,a Bi LSTM-CNN-CRF model that combines the advantages of deep learning and machine learning is constructed.The model can simultaneously capture the feature information between Zhuang characters and words,and combine the conditional random field for joint decoding to improve the recognition effect.In the experiment based on the Zhuang named entity recognition annotation dataset constructed in this paper,its performance is better than other comparison models,and it is suitable for the Zhuang named entity recognition task.3.According to the characteristics of Zhuang named entity boundary,the F-Bi LSTM-CNNCRF model combined with the capitalization feature of Zhuang words is proposed.The F1 value of the experimental results is 2.27% higher than that of the BiLSTM-CNN-CRF model,reaching80.37 %.This shows that adding the features of capitalized Zhuang words to the F-Bi LSTM-CNNCRF model can effectively improve the performance of the model.4.Finally,to realize the proposed algorithm model into practical application,an online Zhuang named entity recognition system is designed.The system provides a visual operation interface,which can perform real-time named entity recognition on the input Zhuang language text.