Intelligent Analysis Of Resident Complaint Cases For Town Management

Matter mapping is an event-based knowledge map that extracts key information from text and establishes links between entities and events to form a structured knowledge base of events.In town management,the use of event mapping technology to analyse residents’ complaints can quickly obtain key facts,clarify the events,provide clues for further investigation and problem solving,help town management to identify the nature and causes of complaints,and quickly respond to and resolve complaints.In this paper,we build a solution for intelligent case analysis of residents’ complaints in town management,including the definition and extraction of events,the definition and identification of similar cases,the extraction of event relationships and the data collection and analysis platform.By analysing the actual case complaints in a town,this paper summarises seven case types,defines the event elements corresponding to each case type and establishes an event analysis dataset for complaint cases.On this basis,a joint event extraction model for the domain is proposed,and end-to-end event extraction at the sentence level is realised;the identification of similar cases in the town is realised through cosine similarity and BP neural network,and event relationship extraction is realised through pattern matching;a complaint data collection and analysis application for town management is designed and realised.The details are as follows:(1)resident complaint case events are defined,and an end-to-end model based on the sentence level is proposed for extracting the event type of the case as well as the event elements from the text.Due to the small size of historical datasets in practical applications,a self-training approach is introduced to learn more knowledge from a small amount of annotated data,improving the effectiveness of the model on small sample datasets.The experimental results show that the proposed model can effectively extract information related to the complaint cases of urban residents from the text with a maximum F1 value of 87.35% with a small amount of annotated data.(2)The criteria for similar cases of urban residents were defined,and a method for judging similar cases was proposed.Through the cosine similarity formula,the calculation of the similarity of case event elements under different event types was achieved,and then the BP neural network was used to judge similar cases.The relationships between urban complaint events are mainly of two types: causal and cis-commitment.Using a pattern matching-based approach,an extraction template for event relationships was designed to realise the analysis of related cases.(3)For town management,a complaint case collection and analysis module was designed and implemented,which has been put into use in the actual system.The main functions include data collection and complaint case analysis.The data collection is responsible for collecting real complaint data from residents,and the complaint case function uses map sdk technology to visualise the analysis results based on matter mapping technology,case similarity and so on.