Research On Agricultural Geological Named Entity Recognition Method Based On Deep Learning

The research work of agricultural geology in China began in the 1950 s,and its research results have promoted the development of agricultural modernization and national economic output in China,opening up a new way for geological work.With the development of the information age,the scale of data continues to expand,and agricultural geological data continues to increase.Obtaining agricultural geological information efficiently and accurately from unstructured text data is one of the difficulties encountered in current agricultural geological data management.Agricultural geological named entity recognition is a technology used to identify agricultural geological entities from a large number of texts.This method provides a solution for agricultural geological information mining and provides a reference for agricultural geological data analysis.However,named entity recognition has many challenges,and the characteristics of various types and forms of agricultural geological named entities have brought difficulties to the research of agricultural geological named entity recognition methods.Therefore,agricultural geological named entity recognition is a very meaningful research topic.This thesis constructs an agricultural geological annotation corpus for training and evaluating agricultural geological named entity recognition models.This thesis compares and analyzes deep learning models from three aspects: pre training,feature extraction,and decoding methods,and proposes an agricultural geological entity recognition method based on deep learning.This method is also used to analyze the research hotspots of agricultural geological papers in the past 10 years.The main research content of this article is as follows:(1)Construct a corpus of agricultural geological annotation.In order to reduce the impact of data on the training of agricultural geological named entity recognition models and improve the recognition effect,this thesis developed four specifications for agricultural geological named entities,entity collection principles,and annotation systems by integrating agricultural geological professional knowledge and the construction method of generic domain datasets,and constructed an agricultural geological annotation corpus for agricultural geological named entity recognition.The connotation of the corpus covers a comprehensive knowledge of agricultural geology.(2)An agricultural geological named entity recognition method based on deep learning is proposed.Through analysis of pre training methods,feature extraction methods,and decoding methods,12 deep learning methods were selected for agricultural geological named entity recognition.The optimal method,namely XLNet-Bi LSTM-CRF,was obtained through analysis of model performance comparison.This method solves the problem of nonindependent prediction by introducing an XLNet model,and converts each word in a sentence into a word vector;Capturing long-distance contextual information in agricultural geological texts by introducing a Bi LSTM model;Finally,CRF is used to ensure the correctness of the tag sequence.The experimental results show that the accuracy of the XLNet-Bi LSTM-CRF model is 91.12%,the recall rate is 91.01%,and the F1-score is91.06%,which is superior to other models and has accurate recognition effects.(3)Text analysis of agricultural geology.Apply XLNet-Bi LSTM-CRF to the agricultural geological related papers collected in this article in the past decade to identify agricultural geological named entities.Compare the recognition results with manual recognition results and traditional named entity recognition results to verify the practicality of the model in practical information recognition.Based on the identified named entity results,various highfrequency entities appeared during the analysis were analyzed to obtain the research hotspots of agricultural geology in recent years.The results were presented through word cloud methods.