Unsupervised Microblog Comment Value Analysis

With Microblog gradually become an important medium for people to obtain and deliver information,it produce huge amounts of microblog and comments every day.It has important application value to find an effective method to exclude spam comments and pick up the valuable comments.These valuable comments can be supplied to the readers,or be provided to tasks like public opinion analysis and text mining.This article conduct some deep research on valuable comments recognition in Chinese microblog.The proposed approach can automatically label high-quality training data in the absence of human intervention,by comparing the correlation of comments and microblog.The main work includes:1)Chinese microblog comment data capture and analysisWe conduct some statistics analysis on these data.Furthermore we select a certain number of comments to read and mark their value,we also analyzes the distribution of the value of these microblog comments.2)Unsupervised microblog comment analysis methodConsidering the features of microblog and its comments,which are very short and include many divergent topics,we presents an unsupervised comment value analysis method,which can automatically find the high quality training data,generate comment classification model belongs to the specific microblog,and then use the model to assess the comment value.The experimental results show that,this method perform well on the task of valuable comments recognition.3)Design and implementation of microblog comment filter system(MCFS)Using the method proposed in this paper,we design and implement the Microblog Comment Filter System(MCFS).The system crawl microblog and its comments data from Sina Microblog,extracts high value comments,and shows them in the form of webpage.