基于知识图谱的Android端农技智能问答系统研究

Design of Agricultural Question Answering System Based on Knowledge Graph

针对目前普及程度较高的以电话直接咨询、集中技术培训和专家现场指导为主的农业信息服务,受时空和人力限制,存在及时性和便捷性欠缺的问题,研究开发Android端农技智能问答机器人APP,为农民提供信息服务。利用爬虫工具采集互联网平台上的海量农技问答数据,经过预处理后形成语料。对语料特征进行自动标注后训练CRF模型识别农技命名实体。并根据词频和信息熵计算命名实体的评价指数,构建"农作物-病虫害-农药"三元组知识库。将知识库导入Neo4j建立农技知识图谱。在Android端集成命名实体识别和知识图谱查询推荐算法,解决用户问题的关键词识别和查询结果的择优推荐问题。所设计问答系统为农技问答提供了一种智能解决方案,具有较高的自动化程度和应用价值。

At present,the popular agricultural information services are mainly telephone direct consultation,centralized technical training and expert on-site guidance in China.Due to the limitation of time and space and manpower,there is a lack of timeliness and convenience.Through the research and development of Android agricultural technology intelligent question answering robot APP,agricultural information service can be provided for farmers.Crawlers were used to collect a large number of agricultural technology Q&A data on Internet platforms,which were preprocessed to form a corpus.The CRF model was trained to recognize the agricultural technology named entity after automatically labeling the corpus features.According to word frequency and information entropy,the evaluation index of named entity was calculated to construct the triple knowledge base of“crops,pests and pesticides”.The knowledge base was imported into Neo4 j to establish the agricultural technology knowledge map.The algorithm of named entity recognition and knowledge map query recommendation was integrated in Android to solve the problem of keyword recognition and query result recommendation.This question answering system can provide a intelligent solution for agricultural technology Q&A,which had a high degree of automation and application value.