Knowledge Graph-Driven Weather Overview Generation for the Beijing 2022 Winter Olympic Games

Weather Overview is regarded as one of the crucial meteorological services supporting the Beijing 2022 Olympic and Paralympic Winter Games (Beijing 2022). As generation of Weather Overview involves multiple-data, large-scale weather conditions, and vulnerability to weather changes, there still exist quite some challenges in obtaining Weather Overview. At present, knowledge graph (KG) is believed to be an effective way to describe information and knowledge. Thus, this study focuses on development of a framework to automatically generate Weather Overview using KG. We first present a three-layer KG model to generate accurate content of Weather Overview: (1) knowledge acquisition of entities and relationships to construct the specific corpora; (2) knowledge representation of the relationships between weather conditions and the events based on ontology; (3) knowledge application of corpora, variables, and weather conditions to query and reason knowledge with Neo4j. Moreover, an XML schema is used to achieve the standardized Weather Overview, which is formed by sentence-paragraph-text generation. This model is validated for a typical case at the Yanqing National Alpine Skiing Centre in Beijing 2022. Compared to the manual method, the accuracy and standardization of Weather Overview can be maintained above 90%, and it can be automatically generated within seconds. The method proposed in this study provides a helpful meteorological service solution to other large-scale sports events.