Construction And Application Of Knowledge Graph For Anhui Tea Based On Knowledge Extraction

Anhui Tea industry,as a traditional advantageous agricultural industry in Anhui,guarantees the economy of mountainous areas in Anhui.But in recent years the development of the Anhui Tea industry has stagnated,in urgent need of high-tech for its secondary empowerment.Now is the digital era,information technology is more and more popular.The combination of agriculture and computer technology can fully excavate agricultural data and develop agriculture more scientifically and efficiently.The construction of knowledge graph based on Anhui Tea helps to quickly mine Anhui Tea knowledge from massive information and process and integrate fragmented information.In the research,the construction of knowledge graph of Anhui Tea based on knowledge extraction is studied in depth,and the specific work is as follows.(1)The research firstly composes and summarizes the current research status of the Anhui Tea industry,knowledge graph and knowledge extraction technology.Then,pre-training models and feature extraction models commonly used in knowledge graph construction techniques and knowledge extraction techniques are summarized.Then the Scrapy crawler framework was used to crawl the text data related to Anhui Tea in tea websites,and the texts related to Anhui Tea in books compiled by experts were screened as experimental data.(2)In order to address the problem of multiple sources and complexity of Anhui Tea data,and the weak ability of traditional shallow sequence annotation model to obtain context-dependent information in the entity recognition stage,the research proposes the ATea-BTC(Anhui Tea-BERT-TENER-CRF)model.Firstly,BERT(Bidirectional Encoder Representation from Transformers)is used to pre-train the datasets and transform characters into low-dimensional dense vectors.Then,TENER(Transformer Encoder for NER)and CRF(Conditional Random Field)are built on BERT to enhance the attention of distance and direction perception by TENER,which solves the problem that it is difficult to distinguish the direction of context by conventional methods.And it improves the performance of boundary recognition to the maximum extent and achieves good entity extraction.(3)For the problem that the Anhui Tea data involves multiple links,complex relationships between entities and insufficient labeled data,the research proposes the ATea-RPA(Anhui Tea-RLRE-PCNN-ATT)model for relation extraction.The sentences are denoised by RLRE(Reinforcement Learning for Relation Extraction)model first,then the relation is extracted by PCNN(Piecewise Convolutional Neural Networks)model of remote supervision.Finally,the performance of relation extraction is optimized to the maximum extent by capturing more fine-grained semantic features through the Attention-based multiple instance learning.(4)Based on the research of entity recognition and relation extraction for knowledge extraction,combined with the operations related to knowledge fusion and knowledge storage,the research completes the construction of the knowledge graph of Anhui Tea.In order to meet the needs of different people,the knowledge graph based on Anhui Tea knowledge question and answer platform is built.The platform can assist tea farmers to scientifically breed tea,assist tea enterprises to make decisions,help consumers to choose tea and promote the culture of Anhui Tea.