Letters Classification Management Based On The Topic Model On The Government Convenience Services

In recent years, local governments increasingly focused on livelihood issues,online convenience services.Because of the convenience and timeliness,it has attracted more and more attention. Many of the city government websites have opened similar functionality to interact with the public or online convenience services.People can be in the form of text interaction with government departments on such a platform, such as Internet users can make their own opinion on an issue and ask the question they don’t know.This network enables the goverment services more transparent,more convenient service to the people and standardization.However,as more and more users on the-site,Also led to the number of letters on that network surge.How to effectively manage these letters have become the current online convenience services development a matter of concern. Some online convenience services just let users fill in the letters in their own classification when so often there are some deficiencies,such as:As classified this operation is not necessary,which means there are a lot of Internet users may not be filled out on your own letter classification; there are many Internet users do not know that will be included in the letter to fill in what kind; it may be because the users have their own cultural level is not high,the letter type misjudgment,which classified the wrong category. Therefore,to find a proper way to train government services for letters of text on the classified management has become very urgent. With the development in recent years based on machine learning and text classification technology matures topic model train makes the letters on government services could be better classification management possible.This dissertation try to train for the government services such applications propose a new kind of topic modeling of automatic text classification method-Thisdissertation is to solve classification problems and a new field because of their characteristic letters text on 12345 train government services,resulting in the previous version of this dissertation can not be taken step by step method for automatic classification.Therefore, the method of this paper is different than the previous text classification are:mail this dissertation due to train government services on a large number of words only in the presence of 150 words or less short text letters.so a) This dissertation will take a topic model and non-letter text vector space model to model.In dealing with the letters found that,because the letters are on government services train people to fill in some city,not only related to the contents of the letter the city’s problems,and has a strong randomness and regional,which led to a lot of randomness the presence of noise words,regional cause there were a lot of letters in the text of regional lexical items,these lexical items are forcibly separated segmentation software or as a stop word is deleted, it is very unreasonable.So,2) This paper proposes a combination of word frequency and document frequency dictionary processing techniques to solve irrational letter text.After a lot of experiments show that the use of the theme of model letters and text classification modeling good results,combined with the two vocabularies processing technology can not only deal with the noise letter word can also improve the classification results in a letter to a certain extent.