Research On The Identification Of Geographical Names Of News Events And The Technology Of MAP Linking

With the advent of the era of big data,news events are presented to users in a massive and diversified form.The organic combination of news events and geographic information provides users with news event information and geographic location data.However,news text itself has the feature of sparse value,and sometimes the mass information does not bring convenience to users,but even causes confusion,so users need a good method of filtering knowledge.When organically combining the news event text information and the geospatial data information to filter the information that is helpful to users,we are faced with the following problems: there are various types of network news text,so it is difficult to filter the text categories that users really need;Chinese place names are difficult to identify in news text because of their particularity;Place-name entities in news events are often presented to users in the form of text,which is not intuitive and simple.In view of the above three problems,this paper mainly studies the identification of Chinese place names from news text and links them to digital maps.This paper first studies the text classification methods based on deep neural network,and compares two mainstream text classification methods,to explore the feasibility and effectiveness of text classification.Labeling and classifying six categories of news text that often involves geographical location,and adding some news text that is hot but does not involve geographic information as a supplement category,which constitutes a total of seven types of news text.By comparing the convolutional neural network and loop neural network model training effect of news text data,this thesis uses convolution neural network on character level to classify text types.Secondly,complicated the Chinese place names are identified.A conditional random field model based on statistics and rules is used to identify Chinese geographical names.On the basis of the model,an algorithm based on basic information entropy and pointwise mutual information is proposed to calculate the word formation rate of a single word in an entity,so as to improve the recognition efficiency and accuracy,which effectively identifies Chinese geographical names.Last,according to the digital map of news text place name positioning,this paper designs a system to recognize and link the place names in the news events,which is based on the VS tool to set up places positioning system.Thus the classified news events in which the place names precisely positioned in the digital map are presented to users intuitively,which facilitates users to effectively make use of the news text.This paper classifies news text based on neural network,which effectively filters the text information and uses the convolution neural network to classify news text.In order to explore the recognition of complex place names in news texts,an algorithm of place names recognition based on information entropy and mutual information of points is proposed,which effectively integrates multiple data information and presents valuable information in news text to users in the form of visualized digital map.The research results can be used for user navigation,disaster avoidance and hot news tracking ect.