Automated Knowledge Extraction in the Field of Wheat Sharp Eyespot Control

Abstract

Wheat sharp eyespot is a soil-borne fungal disease commonly found in wheat areas in China, which can occur throughout the entire reproductive period of wheat and has a great impact on the yield and quality of wheat in China. By constructing a domain ontology for wheat sharp eyespot control and modeling the domain knowledge, we aim to integrate and share the knowledge in the field of wheat sharp eyespot control, which can provide important support and guidance for agricultural decision-making and disease control. In this study, the literature in the field of wheat sharp eyespot control was used as a data source, the KeyBERT keyword extraction algorithm was used to mine the core concepts of the ontology, and the hierarchical relationships among the ontology concepts were extracted through clustering. Based on the constructed ontology of wheat sharp eyespot control, the schema of knowledge extraction was formed, and the knowledge extraction model was trained using the ERNIE 3.0 knowledge enhancement pretraining model. This study proposes a model and algorithm to realize knowledge extraction based on domain ontology, describes the construction method and process framework of wheat sharp eyespot control domain ontology, and details the training and reasoning effect of the knowledge extraction model. The knowledge extraction model constructed in this study for wheat sharp eyespot control contains a more complete conceptual system of wheat sharp eyespot. The F1 value of the model reaches 91.26%, which is a 17.86% improvement compared with the baseline model, and it can satisfy the knowledge extraction needs in the field of wheat sharp eyespot control. This study can provide a reference for domain knowledge extraction and provide strong support for knowledge discovery and downstream applications such as intelligent Q&A and intelligent recommendation in the field of wheat sharp eyespot control.