Research And Implementation Of Text Classification Method For Social Governance

In recent years,in order to better understand the public opinion,government departments have established a hotline for people's livelihood for the public to express their appeals,and it has become an important channel for government departments to understand current social hot spots.With the increasing amount of data,problems such as different standards of classification and low classification efficiency have also been exposed,government can't find the current hot social information efficiently.Therefore,how to solve the shortcomings of the manual classification,quickly and accurately understand the current hot issues that the public are concerned about,and deal with the demands of the public in a timely manner could greatly help the government's management.The text of the people's livelihood service hotline belongs to the category of the short text.It has short content and sparse text features,so the traditional text classification methods are not very good in the performance of the short text.How to optimize the existing traditional classification methods based on the characteristics of the people's livelihood service hotline text,improve the text classification performance,and meet the personalized needs of the people's livelihood service hotline text is a challenge for current research.This paper starts a series of studies on the above issues.The main work can be summarized as:1.Construction of short text classification corpus in the field of people's livelihood.Using the real 12345 people's livelihood service hotline text data to construct a short text classification corpus in the field of people's livelihood.Based on this corpus,a special stopword vocabulary and feature vocabulary are generated for short text processing in the field of people's livelihood.Experiments show that traditional classification methods have good performance and stability in this corpus.2.Research on the short text classification in the field of people's livelihood.On the basis of the traditional classification methods,this paper proposed three short text classification methods based on the characteristics of the short text of the peoples' livelihood service hotline: a classification method based on the feature vocabulary,a classification method based on the synonym vocabulary and a classification method based on the topic models.Experiments show that the improved methods have better classification performance(3%~5%)than the traditional methods.3.The application of automatic text classification in social governance.Based on the short text categorization corpus constructed in this paper and the proposed text classification methods,this paper has built an automatic classification and hotspots mining system for social governance.This system realized the automatic classification of the people's livelihood service hotline text and made data mining and analyzed on the text with the time and space.Also,this system displayed the data mining results through the visualization techniques,and it could help the government's managers to focus on the social hot spots...