Sentiment Analysis Approach Forpublic Comments On The Internet

With the rapid development of network technology,we have entered the era of big data.In the environment of a universal network,people access the Internet easily and can make a personal evaluation of the interesting object(such as news,goods and people's livelihood of personal opinion)on the web at anytime.Thus,the scale of comments which are posted by people becomes large.However,Users will spend extra time and have a bad experience when they can't get the comment in which they are interested at once.Therefore,it is necessary to extract the important information from the comments in an appropriate way so that users can quickly locate.User usually expresses in his comments.So we can recognize emotions in comments and classify these comments,which help users to retrieve and analyze information.This paper presents a classic method based on the classification of emotional dictionary,specific steps are: 1)Crawl by the crawling network technology to save comments;2)Preprocess text and removing stop noise data;3)Build emotional dictionary.However,due to the inherent shortcoming of the classification method based on the emotional dictionary,and the features of public comment multi-source,such as huge amounts of data,the shorter length,variety of forms,informative,intensely emotional characteristics,this paper also proposes a sentiment classification method for Internet public short comments based on the Convolutional Neural Network(CNN).Compared with the traditional method of emotion classification,CNN has a unique advantage in the extraction of local features and has two major characteristics:one is the local perception,the other is parameter sharing.Through these two characteristics,we are enabled to greatly reduce the training parameters in the training process.We design an eight layers of CNN structure,which is an emotion classifier for public comments and makes it more accurate to extract local features.The method is as follows: 1)Use Word2 Vec to convert the comment text into the form of spatial vector of the word;2)Convert the two-dimensional matrix text into gray picture format;3)Input the picture data into the emotion classifier based on CNN for training and learning.4)Recognize the emotion in short comments by sentiment classifier.From the experimental results,we find that the method based on CNN is superior tothe classical method in the accuracy rate,which proves the feasibility and effectiveness of our idea.