Research Of Aspect Level Sentiment Analysis Based On Deep Learning

With the rise of social networks,more and more people like to share their opinions on the Internet,and this sort of text with individual’s subjective attitude has great value for further exploration.The primary purpose of the Aspect Level Sentiment Analysis is to determine the textual sentiment from a specific target’s perspective.The traditional Sentence Level Sentiment Analysis can only calculate the sentiment value of a complete sentence,which however,with the increasing demand of users,fails to accurately describe what the user wants to know.Therefore,the level of analysis needs further refinement.From the perspective of sentiment classification,the analytical approach should be able to capture any significant information in the context.As the rapid development of deep learning,it provides the Aspect Level Sentiment Analysis a brand new method.This paper aims to complete the tasks of the Aspect Level Sentiment Analysis based on deep learning model.First of all,this paper studies the structure and the algorithm of the Convolutional Neural Network model,including its input processing,pattern training,and overfitting preventing method.And then this paper puts forward the text pretreatment and the text representation method,building framework of sentiment analysis based on Convolutional Neural Network algorithm.Through controlled experiments on open data set,the deep learning model introduced in this paper contributes to a higher accuracy than traditional machine learning classifier Naive Bayes and Support Vector Machine do,verifying the effectiveness of deep learning framework in textual sentiment analysis.Then,through the features of the Aspect Level Sentiment Analysis,this paper puts forward the framework of the AttenNet model,which contains multiple computing layers and a memory module.This framework is able to capture important information of the aspect words in context.After that,a series of Aspect Level Sentiment Analysis experiments are conducted on open data sets,proving the AttenNet model introduced in this paper has a better forecasting effect and a faster processing speed than the five basic algorithms.Last but not least,this paper argues that the accuracy increases with layers by comparing models with distinct computing layers.