Research On Semantic Mining Of Agriculture Science And Technology Information Based On Deep Learning

At present,there is a widespread ‘abundant data and poor knowledge' phenomenon in agricultural literature,which shows the amount of data in agricultural specialized literature has increased rapidly,and information overload has become severe.And users urgently need to compress overloaded agricultural information into key and effective knowledge in order to utilize efficiently.However,traditional information extraction techniques can lead to the lack of semantic semantics,semantic ambiguity,and semantic incoherence after extraction of overloaded information,which can not meet the increasing accuracy,relevance,and integrity requirements of agricultural users.In view of this situation,the paper studies related semantic mining technology theory and grasps the requirements of agricultural data application scenarios.First,combining the deep learning methods requires the characteristics of massive information to train hyperparameters,we research on automatic summarization methods based on deep learning and generative,to increase the utilization of existing large-scale agricultural science and technology information;Second,in order to improve traditional information extraction methods,we use deep neural network technology to train and learn the characteristics of agricultural science and technology information,and provide new methods and technologies for information acquisition and analysis in the field of agricultural science and technology information;Third,in order to improve the accuracy of the search results,to save the cost of human marking,the paper works on the development of a new automatic extraction system for agricultural science and technology text information abstracts.The paper takes agricultural science and technology literature information as the research object.Without the need of artificial characteristics processing and specific domain knowledge,in order to improve the utilization rate of massive agricultural science and technology information,from the perspective of word embedding instead of the traditional bag of words,research based on word embedding self-encoding network,which represents the internal semantic logic of agricultural science abstracts data,it can model agricultural scientific literature data on a large scale.At the same time,in order to solve the semantic problems existing in the traditional information extraction technology,the method with Seq2 Seq based on recurrent neural network combined with the Attention mechanism is studied to effectively obtain the overall sequence characteristics of agricultural science and technology information,which is by training the generated automatic abstract model to explore new semantic mining methods that are applicable.In detail,based on the related theories of deep learning in natural language processing,the paper studies the theoretical basis of semantic mining of agricultural science and technology information,carries out the preprocessing of agricultural science abstracts data sets,and mainly selects Seq2 Seq and Attention mechanism as semantic mining methods and trains the automatic abstract generation model.Based on the ROUGE evaluation index system of the American Institute of Scientific Information,the model was qualitatively evaluated and quantitatively evaluated,and the deep neural network hyper-parameters in the field of agricultural science and technology information were optimized.Finally,in order to persist the results of the automatic abstract model,a semantic and mining automatic abstract extraction system prototype based on B/S structure and user interface were developed to meet the needs and application scenarios of agricultural science and technology.Through the study of this paper,the deep learning technology is applied to the field of agricultural text information generation and retrieval,trying to make up for the shortcomings of the traditional extractive automatic summarization method.Massive information "compression" experience has been gained.At the same time,using the current mainstream deep learning technology and related tool sets,related topics in the text are automatically extracted,and information extraction system prototypes and user interface interfaces for information retrieval and analysis of agricultural science and technology are developed,which improves the service level of agricultural information acquisition and analysis.