Design And Implementation Of Information Extraction System For Internet Meteorological Disasters Texts

In recent years,all kinds of meteorological disasters have occurred frequently,bringing huge loss of life and property to the people in disaster areas.China has a vast land area and a variety of climate types,including the monsoon climate in the southeast coastal area and the continental climate in the northwest inland area,and the temperature zone from the northernmost cold temperate zone to the southernmost tropical zone.China is one of the countries most seriously affected by meteorological disasters in the world.Meteorological disasters accompany the whole process of the development of human society and have a great impact on people’s production and life.In order to reduce the loss of life and property caused by meteorological disasters to the people as much as possible,it is very important to take timely and accurate measures.With the rise and rapid development of Internet technology,more and more information can be quickly spread through the Internet.Through the Internet,people can get a lot of instant information from various websites and social platforms,and can quickly learn the occurrence and progress of various emergencies.Although there are a large number of meteorological disaster texts on the Internet,these texts are all unstructured text information,and it takes a lot of time and manpower to extract important information related to meteorological disaster manually.Therefore,it is of great significance to extract important information in Internet meteorological disaster texts efficiently.In view of the characteristics of meteorological disaster and Internet meteorological disaster-related texts,this paper designed and implemented an information extraction system for Internet meteorological disaster texts.The main work is as follows:In this paper,a large number of Internet meteorological disaster texts are obtained from news websites through web crawlers,and they are manually annotated,and a meteorological disaster texts data set which can be used for entity recognition and entity relationship extraction is constructed.This paper compares the performance of BiLSTM based on char level and lexical features,and compares the performance of BiLSTM based on char level and BERT based on char level,the bert-based model has the best performance.At the same time,the dictionary and data enhancement are combined to improve the model performance.This paper explores the realization and optimization method of document level entity relation extraction of meteorological disaster text.The effects of the number of contexts,different text representation methods and additional features on the performance of entity relationship extraction are compared.This paper designs and implements an information extraction system for Internet meteorological disaster text,and tests each function module.The test results show that each function of the system runs well and meets the requirements of use.