Study And Implementation Of Sentiment Analysis For Short Chinese Text

With the continuous development and popularization of the Internet in China,more and more users choose to publish their opinions about the products,policies and emergencies on the network.In recent years,the monitoring of public opinion information has become an important means for the government and enterprises to gain public opinions.Analyzing the emotional polarity of public opinion information is one of the key problems in the field of public opinion monitoring.At present,most of the mainstream Chinese emotion analysis methods are based on the features of the words,but the users have short text and incomplete grammatical structure in the comment information published on the network,which results in the lower accuracy of the Chinese word segmentation,which further affect emotional analysis results.In view of the significance of the above research and the existing problems,this article carries on the Chinese emotion analysis and the public opinion monitoring aspect research,the main work content includes:(1)Combined with Naive Bayesian,support vector machine,random forest and convolutional neural network,we propose and implement a method based on character to analyze sentiment of Chinese text.(2)On the open data set,we train and test different models and algorithms under different combinations of parameters and different parameters.The results show that models based on character features has a better accuracy than models based on word characters,under the same conditions.(3)According to the demand of public opinion monitoring and algorithm research,combining with research results and mainstream technology solutions,a public opinion monitoring and experiment platform is designed and implemented.