Research And Implementation Of Tea Knowledge Question Answering System Based On Knowledge Graph

Services of agricultural information are an important part of the construction of "digital villages",It is very important to promote the construction of "digital villages".How to use the open data and artificial intelligence technology to realize the corresponding information system on the Internet and provide services of agricultural information for agricultural related people is an important task.The agricultural knowledge question and answer system is an important way of services of agricultural information.Tea is an important economic crop in China.Constructing a knowledge question and answer system in the tea field not only helping farmers of tea and technicians to get relevant professional knowledge,but also providing decision-making for government.The thesis conducts an in-depth study on the tea knowledge question and answer system based on the knowledge graph,and uses the tea text data of the National Agricultural Science Data Sharing Center and China Tea website to construct dataset for joint extraction and knowledge graph completion.This thesis mainly research content is as follows.(1)Aiming at the problem of polysemy and overlapping relations of tea text,a joint entity-relation extraction model is proposed,which uses the BERT pre-training language model to fine-tune the word embedding using context,and avoids polysemy in the text.On the basis of the table representation method,the last character matching algorithm is proposed.It avoids the overlapping relations.The experimental results show that this model has a better performance of joint extraction compared with other joint extraction models such as Bi-RNNCRF.(2)Aiming at the problem of incomplete of tea knowledge graph,knowledge representation learning model based on the name information of triples is proposed.The traditional knowledge graph completion model only considers the information of topological structure of knowledge graph without the additional information,which will cause the loss of semantic information and reduce the completion performance.The knowledge representation learning model based on name information of triple,triples as complete sentences in the real world,and then uses all the above sentences to form a training set for word2 vec to train word vectors,so that the entities and relations vectors in the model contain semantics information.The experimental results show that,compared with other models such as Trans E,this model has a better performance of completion of the knowledge graph of tea.(3)Design and implement a tea knowledge question and answer system based on the knowledge graph.The system includes tea knowledge question and answer module,tea knowledge encyclopedia module,experts module,agricultural express module and user management module.The tea knowledge question and answer module can provide effective decision-making for tea enterprises,governments and tea farmers.The tea knowledge encyclopedia module can provide knowledge of tea.The expert team module provides contact information of expert.The agriculture express module displays the latest agricultural news.The user management module manages and maintains user information.