关键词: 深度学习, 知识图谱, 玉米生产过程, 玉米病虫害, 玉米品种, 识别模型, 数据采集

Abstract: Knowledge graph is a graph database that stores knowledge in a structured form. Aiming at the problem that corn growers cannot efficiently and quickly obtain the knowledge of corn production process, a knowledge graph of corn production process based on deep learning was constructed. The semi‑structured data comes from the maize high‑yield cultivation plan in the form of a table formulated with the help of experts in the field of agriculture; the unstructured data acquisition is based on the OCR technology to convert the book into a text document in TXT format. The schema layer was constructed based on the different characteristics of the corn field data, and it was determined that corn varieties contained 1 type of entity and 12 types of attributes, and corn diseases and corn pests contained 6 types of entities and 5 types of relationships, respectively, to form a conceptual layer of corn cultivation knowledge entities and relationships. During the construction of the data layer, the BERT—BiLSTM—CRF model was used to identify entities in unstructured data through the BIOES annotation method. The results showed that the BERT—BiLSTM—CRF named entity recognition model based on BERT—BiLSTM—CRF improved 14.31%, 7.36% and 3.86% in F1 value compared to LSTM, LSTM—CRF, and BiLSTM—CRF, respectively. The precision rate, recall rate and F1 value reached 89.31%, 88.54% and 88.92%, respectively. The completed knowledge graph of corn production process can improve the efficiency of users in acquiring corn planting knowledge and improve the management level of corn planting.

Key words:  , deep learning, knowledge graph, maize production process, corn diseases and insect pests, maize variety, recognition model, data collection