Research On Intelligent Distribution For Police Alarms

Receiving and handling the police alarms is an important responsibility of the public security organ,and the timely and accurate handling after receiving the police alarms is related to the people’s sense of security and satisfaction.Because the description of address information and alarm content in the traditional 110 alarm reception is not accurate enough,there are problems such as high labor cost,long circulation time and low alarm distribution accuracy when manually processing police alarms and matching the police units.These problems seriously affect the quality of the 110 alarm reception and handling work.And they are the difficulty and pain point of the 110 alarm reception and handling mechanism.The named entity recognition(NER)technology and text classification technology based on deep learning could help the front-line police to automatically extract key entities from the alarm records,classify the cases and match the police units according to the addresses,which have crucial value in the actual work of the public security.Focusing on the optimization of police alarm distribution,this paper prominently studied the application of NER technology in the automatic extraction of key entities from police records,and the application of text classification technology in the case classification and police unit matching tasks.The specific work contents are as follows:Firstly,focusing on the surge of unstructured police alarms and the difficulty of entity abstraction,an automatic extraction technology of key entities from police alarm records based on deep learning was proposed.This technology improved the mainstream entity recognition framework and introduced the pre-training language model and multi-head self-attention mechanism.After experimental verification,this technology has better comprehensive performance in speed and accuracy,and could be applied to key entities extraction in public security practical work.Secondly,for the current situation that a large amount of mixed texts of different lengths in police alarms,a general text classification technology that combined attention and cropping mechanism was proposed.This technology combined convolutional and recurrent neural networks to extract local and global semantic information,constructed a dual channel attention mechanism(DCATT)to enhance key areas,and designed a long text cropping mechanism(LTCM)to filter out critical text information.Verified through experiments,this technology could achieve rapid classification of texts of different lengths on public news classification datasets,case classification dataset and police unit matching dataset,and could be utilized for intelligent distribution of police alarms in public security practical work.Ultimately,combined with the practical needs of public security,an intelligent police alarm distribution system was constructed by combining key entities automatic extraction technology and general text classification technology,which includes police alarm input,police alarm entities extraction,case classification,police unit matching and police alarm distribution.This system provides two modes: real-time processing of single police alarm and batch police alarms processing.And the practical application in public security organs have shown that this system could significantly improve the efficiency of police alarms handling.