Location Information Extraction And Spatialization In Traffic Microblog Text

There are various types of traffic information in microblogs.With the improvement of microblog users and publishing convenience,the real-time traffic information released and shared by public security traffic control will increase synchronously.Timely and effective road condition information is an important data base of intelligent transportation service,in which location information is an important component.The location information,such as geographical elements and spatial relations,can be quickly identified and extracted from the rapidly updated traffic microblog text in a timely and effective manner,and then transformed into structured information and fused with road network vector data for spatial positioning,thus realizing the spatialization of text to map,which can be used as a useful data supplement for traditional traffic information collection means and intelligent transportation service.Aiming at the extraction and spatialization of location information in traffic microblog text,the main work and achievements of this paper are reflected in the following aspects:(1)location information extraction from traffic microblog text.According to the syntactic expression features of position information in traffic texts,based on linear reference model,the parts of speech of spatial feature words and their role attributes are extracted,and a position expression mode is constructed and expressed as a Trie tree structure.At the same time,combined with the words and text word object sets of parts of speech obtained from microblog text preprocessing,the mapping with each element in finite state machine(FSM)is established,and the position information in microblog text is recognized and extracted by FSM-based position information extraction algorithm.Finally,9799 microblog texts of urban road conditions in Nanchang,Guangzhou and Shenzhen are taken as experimental data to verify the model.The results show that the accuracy and recall rate of the model are over 85%,which can effectively extract the location information in traffic microblog text.(2)Subject classification of traffic microblog text.Randomly select 1000 microblog texts and divide them into training set and test set according to 4: 1.Firstly,the traffic microblog text is de-noised,and after word segmentation and stopword removal,a corpus dictionary is constructed based on the word bag model,and TF-IDF feature extraction is carried out to transform the traffic microblog text into a training corpus of vector representation,so as to train the LDA model of three types of traffic topics,such as slow traffic,traffic accidents and traffic control.Finally,the trained model is used to predict the topic distribution of the text to be classified,so as to complete the classification of traffic microblog text information.(3)Spatialization of traffic microblog text location information.Taking Shenzhen as an example,firstly,the road network topology is established by using OSM road network data,and the database of road intersection and road section is generated.Then,the initial positioning of location information is divided into three positioning modes: road entrance,road section corresponding to start and end point and road combined with POI.After the text location information is accurately positioned to a certain intersection or road section in combination with spatial relationship,symbols are visualized according to the classification of traffic microblogs,and information such as the type,location and direction of traffic events is visually displayed.