Research On Cereal Crops Knowledge Question Answering System Based On Knowledge Graph

In recent years,China’s agricultural informatization construction has been continuously advanced,which has greatly improved agricultural efficiency and increased farmers’ income.However,due to the incomplete information construction in the field of cereal crops,the vast amount of cereal crop knowledge on the Internet is mostly unstructured or semi-structured,it is difficult for farmers,experts and other related practitioners to obtain the required information from traditional search engines in a timely and accurate manner.The question answering system based on knowledge graph can transform the fragmented cereal crop knowledge in the network into structured knowledge storage,and return accurate answers to users,which can well solve the above problems.Therefore,it is of great significance to study the question answering system for cereal crops knowledge based on knowledge graph to promote the development of informatization in the field of cereal crops.Based on the key technologies of knowledge graph and question answering system,the research studies and implements a question answering system based on knowledge graph for cereal crops knowledge.The main research contents are as follows:(1)Aiming at the problem that most of the current knowledge graphs only contain text single-modal information and the image information is difficult to represent in the knowledge graph,the YOLOv5 model is used for image entity recognition,and the image entity naming label is obtained,in the form of < image entity naming label,image,image URL> triplet,it is added to the cereal crop text knowledge graph,and a cereal crop knowledge graph containing image and text multimodal data is constructed as the knowledge base of the question answering system.(2)Realize the intelligent question answering of cereal crops knowledge based on knowledge graph,and use entity recognition and question classification to realize the task of semantic parsing of cereal crops knowledge questions.The research constructs a question classification model based on Ro BERTa + Bi-GRU + Capsule Network-Attention(RBGCNATT)and a cereal crop knowledge question dataset.Firstly,the R-BGCNATT model uses the Ro BERTa pre-training model to replace the traditional word embedding algorithm Word2 vec to improve the semantic representation ability of words,then introduces two parallel neural network layers in the feature extraction layer: Bi-GRU layer and Capsule Network layer,and an attention mechanism is introduced in the Capsule Network layer to weight the keyword word vectors of the question sentence,the global features and local features of the question are extracted respectively,and the semantic information of the question is more abundantly represented.The R-BGCNATT question classification model is tested on three datasets,and the results show that the model achieves an average accuracy of92.94% and an average F1 value of 92.59% on the three datasets.(3)Design and implementation of a question answering system for cereal crops knowledge based on knowledge graph.Based on the above research content,using the VUE framework,four web page modules are constructed: multimodal data entity recognition module,user question classification module,cypher query statement module and answer generation module.The cereal crop question and answer system is implemented,which meets the user’s demand for question and answer of cereal knowledge.