Research On Entity Relationship Extraction In The Field Of Crop Pests And Diseases Based On Deep Learning

As an important factor restricting agricultural development,crop pests and diseases have always been the focus of attention.At present,a large number of data related to crop pests and diseases have emerged on the network.Through traditional search engines,users cannot efficiently and accurately obtain information on crop pests and diseases,and the application of intelligent pests and diseases such as crop pests and diseases has become a trend of smart agriculture.Therefore,this dissertation conducts research on the extraction of crop pests and related entity relationships,and provides theoretical basis and feasibility basis for constructing knowledge and answer questions about crop pests and diseases.As the basic task of natural language processing,relation extraction is compared with the traditional entity relationship extraction method.This dissertation uses the attention mechanism to combine the four deep learning models of PCNN,CNN,RNN and BiRNN to realize the relationship between entities in the field of crop pests and diseases.The specific work of this dissertation is as follows:1.Completed experimental data collection and preprocessing.Through the study of reptile technology,using Python-based reptile tools,Baidu Encyclopedia as the experimental data source,to obtain data related to crop pests and diseases required for experiments.Use the Jieba tool to complete the word segmentation,part-of-speech tagging,etc.,and use the Word2Vec tool to complete the word vector training to generate the data set needed for subsequent research.2.Realized the identification of the crop pest related entities and the construction of the entity relationship set.The combination of the dictionary and the CRF model is used to identify crop pests and related entities.With the assistance of the plant protection students,the relationship between the entity pairs is determined,the entity relationship set in the field of crop pests and diseases is constructed,and the relationship mark of the entity pairs is completed.3.Research on entity relationship extraction based on deep learning was carried out.In view of the shortcomings of the traditional relation extraction method,the four deep learning models of PCNN,CNN,RNN and BiRNN are used to study the entity relationship extraction in the field of crop pests and diseases.Through the analysis of the experimental results,in the experimental environment and data of this dissertation,the accuracy of CNN extraction is slightly higher than that of PCNN,and the effect is best.Then the entity relationship extraction based on the word level attention mechanism and the four models is adopted.The experimental results show that the neural network model based on the attention mechanism has different effects in the extraction of the entity relationship.PCNN model The accuracy rate is improved by 2.62%,and the extraction effect is more ideal.This dissertation is based on the deep learning model to extract the relationship between entities in the field of crop pests and diseases,which is of great significance,and plays a theoretical role in the construction of the knowledge and answer system for crop pests and diseases.