Research on Meteorological Early Warning Error Correction Model Based on Seq2Seq Deep Learning Method
An error correction model of early warning information is developed for semantic errors in the release of national meteorological early warning information.By establishing a national meteorological historical warning information corpus,training an error correction model based on Seq2Seq deep learning method,and verifying it with a rule model based on statistical methods,a quality control platform for monitoring the legitimacy of warning and forecasting information was formed.A quality control business process of"intelligent semantic analysis+manual verification"was constructed to achieve rapid localization and reminder of sensitive words.Since the early warning quality control platform began to be applied,the error rate of information content decreased by 70%compared with the previous year,and the semantic error correction effect was remarkable.
early warning releasesemantic analysisSeq2Seq deep learningearly warning legality monitoringquality control model
NETL
NSTL
万方数据
