Chinese Text Sentiment Analysis For University's Public Opinion

Due to the rapid development of Internet technology,more and more users like to publish their opinions and comment information on social media such as Weibo,and the network public opinion data has also increased rapidly.It is of great value to analyze and process these data.Because college students are the main force in the Internet army,and they are more active than other groups,the amount of public opinion information in universities is larger and the spread speed is faster,which makes their management more difficult.And the public opinion crisis of universities has occurred frequently,which has caused serious negative impacts on universities.The sentiment analysis of university's public opinion information can help college leaders and student workers better understand the general public's views and opinions on the school,so as to better prevent and respond to these crises.Therefore,mining the text data of college public opinion on some hot issues and analyzing the sentiment tendency of it is of great significance to ensure the harmonious and stable development of colleges and universities.At present,deep learning methods are the most popular research direction in the task of text sentiment analysis.However,the existing deep learning methods do not make full use of the features of Chinese text.The accuracy rate needs to be further improved.On the other hand,CNN-based neural networks usually use pooling to extract data features but cause information loss,while capsule networks use capsule vectors and routing mechanisms to overcome this problem,and can also express the positional hierarchical relationship between local features.Therefore,this paper combines capsule network and CNN to extract local feature information.At the same time,the capsule network cannot selectively focus on important words in the text,and its routing mechanism needs to be improved.The research content of this article includes the following aspects:(1)In recent years,some hot events in Colleges and universities,such as Zhai Tianlin's academic events,were selected to retrieve and screen 10000 Weibo comment data from Weibo,and then the emotional polarity annotation was carried out to create a public opinion data set in university's public opinion.(2)This paper proposes a text emotion classification method based on morphological features of Chinese characters and How Net.Firstly,cw2 vec model and sat model are used to train the word vectors.Secondly,these two different word vectors are used as the input of the two channels of the method.Then,the attention mechanism and convolutional neural network are combined to extract and splice the features of the two channels.Finally,the classification results are output.The experimental results on the two datasets show that the proposed model is superior to many kinds of sentiment analysis methods in terms of classification performance,which verifies the significant advantages of the model in Chinese datasets,and it also verifies the superiority of this method for analyzing the propensity of public opinion texts in colleges and universities.(3)This paper proposes a text sentiment analysis method based on attention capsule network.First,we still use cw2 vec model and sat model to train the word vector and use it as the input of the two channels of the method,then combine the attention mechanism and convolutional neural network to extract the features of the two channels,and then input them into the capsule network to realize sentiment classification.A static routing mechanism is also used,which has a higher accuracy rate than the dynamic routing mechanism.The validity of the model is verified on the public opinion data set of colleges and universities,and it is obviously superior to many kinds of emotion analysis methods.