Classifiction And Recognition Of Chinese Organization Name Based On Deep Learning

In recent years,the research on Chinese organization named entity recognition has made great progress since its development,compared with person name entity recognition and Geographical name entity recognition,this is the main task of Chinese named entity recognition.because of the complexity,randomness and dynamic change of Chinese institution names,the research still can't reach the need of fine granularity division of Chinese institutions.The classification of Chinese organization structure according to the subordinate hierarchy is an important Composition to extract the institutional relationship and construct the knowledge graph of the organization,which has an important influence on the follow-up knowledge mining project,but the relevant research has not been paid much attention in the academic field,and it is seldom explored in the domain of Chinese name entity recognition.How to accurately identify the names of Chinese organizations and hierarchically resolve organizations classification according to subordinate levels,and breaking through the bottleneck of Chinese organization knowledge graph construction are the main research content of the article.This topic starts with the study of the development history and basic methods of the Chinese organization named entity recognition,and then discusses the recognition of the Chinese institution and the classification of the Chinese institution name at the present stage.In this paper,we not only studied the use of different traditional machine learning methods to solve this task,but also studied how to accurately and efficiently classify the names of Chinese institutions using the sequence tagging model based on deep neural networks.The experimental results show that the optimized CNN model is superior to the traditional machine learning model(Decision Tree,MLP,Linear SGD,SVM)and the improved LSTM model,and the accuracy of classification recognition is as high as 75.4The research content and innovations of the thesis are summarized in the following two aspects:1.This topic implements how to identify and classify multi-level(up to 10 levels)organization names based on the recognition of Chinese institution names.At the same time,traditional organization name recognition is done in plain text.The data of this task is limited to the address list,and there is no context information.In addition,the traditional organization name recognition task is relatively independent,regardless of the relationship between the organization name.this task needs to extract the affiliation relationship between agencies.Based on the above considerations,this paper proposes a hierarchical recognition method for Chinese institution names,including an annotation method for hierarchical organization,the construction of data sets based on Chinese character vectors,and the construction of hierarchical annotation model frameworks.2.In order to accurately refine the organizational hierarchy relationship,in the hierarchical division prediction part,the paper not only used the traditional machine learning model(Decision Tree,MLP,Linear SGD,SVM)but also the mainstream neural network model at the present stage.Chinese organization names are expressed as low-latitude Chinese character vectors with the deep neural network.It replaces the traditional artificial feature template for high efficiency.Then it uses different deep neural networks including long-short-term memory neural networks and convolutional neural networks for sequence labeling.through the NNLM-based word vector model combined with deep neural network,the target task which the Implementation of Hierarchical Division in Chinese Institutions was achieved.