Research And Application Of Intelligent Agriculture Question Answering System Based On Knowledge Graph

With the development of agricultural informatization,the agricultural data has grown exponentially,but the organization and representation of these data are not the same,and the correlation between the data is difficult to mine.It takes a lot of time for users to query agricultural knowledge,and it also puts forward higher requirements for users’ own knowledge reserves.How to extract semantic information from the breadth and depth of big data in the agricultural field,summarize and organize it.This paper combines knowledge graph and automatic question answering related technologies to achieve high-quality integration and application of agricultural knowledge.The main research contents of this paper are as follows:1.By analyzing the agricultural field,this paper completes the construction of the agricultural field ontology,including crops,climates,locations,people,functions and other entities and their relationships.2.In the process of entity recognition,based on the pre-training model BERT,the effect of using character sequences as input when training Chinese corpus for BERT is not ideal,and the mainstream vocabulary enhancement methods to improve performance do not inject vocabulary information into BERT In the case of the model,an improved algorithm for the BERT-BiLSTM-CRF model is proposed.and the model in this paper is used to complete the extraction of agricultural vertical entities.3.Build a knowledge graph in the agricultural field.This paper analyzes the multi-source heterogeneous data format,uses the crawler technology to complete the data collection,and proposes a knowledge fusion method of attribute weighted aggregation,uses the Neo4j graph database to store the knowledge graph.4.Complete the agricultural automatic question answering system.Based on the constructed agricultural knowledge graph,this paper analyzes the advantages and disadvantages of template matching and semantic parsing methods.In order to improve the generalization and robustness of question answering,an agricultural automatic question answering system based on representation learning and BERT pretraining model is proposed.5.Complete the knowledge application platform in the agricultural field.It designs and implements multiple functional applications of knowledge query,entity recognition,and automatic question answering.