Sentiment Analysis Of Uyghur Text

Nowdays,the amount of data on social media on the Internet has increased substantially.More and more netizens have become providers of data.Weibo data,forum data,and e-commerce website comment data are all provided by the network democracy.These data contain the attitude of netizens towards a social problem or a product.The analysis of these comments with the network’s democratic outlook has not only economic value,but also social value.Researchers are trying to extract useful information from these sentiment data to help managers make better judgments.Text sentiment analysis is the use of computer to automatically make emotional judgments on these text data.This paper studies the sentiment analysis of Uyghur text.Compared with Chinese and English text sentiment analysis,the Uyghur text sentiment analysis started late.Although previous researchers have conducted some research on Uyghur text sentiment analysis,they are all models for single machine learning or deep learning.Text sentiment analysis generally consists of text representation,feature selection,feature extraction,and classification algorithm selection.Feature extraction technology is different from feature selection technology.Feature extraction is the transformation of original features into features that improve the classification effect by some algorithm.Feature selection is to select more representative features from the features.This paper proposes two methods of feature extraction: text feature extraction based on feature selection and deep belief network and text feature extraction based on convolution recursive depth model.Text feature extraction based on feature selection and deep belief network solves DBN consumption.Time and calculation of expensive problems,feature selection to delete some useless features,thus reducing the input dimension of DBN network;text feature extraction based on convolution recursive depth model solves the problem that convolutional neural network can not capture text long-term dependence The problem of relationship.Both of these methods have shown good results in the Uyghur textsentiment classification.Text feature extraction based on feature selection and deep belief network is better than deep learning algorithm(convolution neural network,cyclic neural network,deep belief network)in training time,and text feature extraction based on convolution recursion depth model is higher in accuracy than Naive Bayes,support vector machine,convolutional neural network,and recurrent neural network.The first algorithm proposed in this paper combines machine learning with deep learning.The second algorithm combines two deep learning models to avoid the shortcomings of a single model.