Research on targeted service technology for early warning information of meteorological disaster based on NLP

Abstract

Affected by global climate change and the superposition of urban construction, the climate of city meteorological guarantee service puts forward new requirements and new challenges. As the first line of defense of meteorological disaster prevention and reduction, it is very important to realize the refinement targeted service of early warning information. There are some problems in the traditional early warning information service, such as the difficulty of accurate service of early warning information, the lack of precision, and the insufficient mining of early warning text information. This paper mainly analyzes the text description characteristics of early warning information of meteorological disaster, constructs the early warning information knowledge extraction process, constructs the early warning information labeling system, and realizes the early warning effective time extraction method based on conditional random field model, Early warning affected areas extraction method based on bidirectional long-term and short-term memory neural network and early warning cautions extraction method based on bootstrapping weak supervised learning method. Finally, taking the early warning information targeting service of meteorological information decision support system as an example, this paper tests the early warning information extraction methods, and preliminarily realizes the early warning precision targeting service in the decision support service of meteorological disaster prevention and reduction.