Emotion Analysis In Natural Language Processing Based On Eye Tracker

The eyes are the Windows of the soul.Most of the information received from the outside world comes from the eyes.At the same time,people also release some potential information through the eyes.Eye tracker is an instrument that integrates infrared light source and professional camera to obtain eye movement data.Through the eye tracker,such as eye fixation point,eye movement trajectory,eye fixation time and other data can be obtained.Experts and scholars in various fields around the world use eye movement data for further analysis and research.Natural language processing is an important research direction in the field of artificial intelligence.It is a comprehensive discipline that studies how to make computers better recognize natural language,understand natural language,and use natural language to communicate with human beings.Natural language processing research is applied in many fields,including machine translation,sentiment analysis,automatic question and answer,speech recognition and other current hot areas.Text Sentiment Analysis(TSA)is a process of analyzing and processing the subjective text with emotional color by using natural language processing technology and text mining technology.At present,the research of text sentiment analysis covers many fields,including natural language processing,text mining,information retrieval,information extraction,machine learning and so on.It has attracted the attention of many scholars and research institutions.In recent years,it continues to become one of the hot issues in the field of natural language processing and text mining.On the basis of the existing application scenarios of eye tracker,this paper proposes the research of natural language sentiment analysis based on eye tracker.The study was able to analyze participants' emotional preferences for the aspects of the information by viewing it for a short period of time.The specific work content and research results of this paper include the following aspects:Emotion classification algorithm selection:based on emotion dictionary emotion analysis and logistic regression based on machine learning algorithm and support vector machine algorithm experiment respectively,in the local data collection of training and testing,support vector machine(SVM)algorithm because it does not need too much prior knowledge and higher classification accuracy,as an emotional classification algorithm in this paper.Word vector training model:Based on the combination of two language models,CBOW and skip-gram,and two computational optimization methods,hierarchical Softmax and negative sampling,four different word vector training models are established.By comparing the training time of the four models and the advantages and disadvantages of the output word vector,CBOW language model and negative sampling optimization method with less time to select and better word vector effect are determined.Then the dimension of the word vector,the minimum threshold of the word frequency and the size of the context window in the model are adjusted by the control variable method to determine the best training model of the word vector.Feature extraction methods:based on word frequency and word frequency matrix inverse text vector design,a weighted term vectors as the weight of word to TF-IDF matrix,combining with term vectors,the effectiveness of this method is proved by Python,and with the two methods respectively as input,comparing the classification effect of this method is proved by the results of classification effect has large improvement.Obtain and process eye movement data with eye tracker:build an experimental platform of eye tracker,collect the line of sight information of the tester when viewing test pictures with eye tracker,and obtain the text of interest in the line of sight by obtaining coordinate set and analyzing and processing effective coordinate set.After the text is preprocessed,the weighted word vector proposed in this study is used as the input,and the emotional polarity of the text is output by the support vector machine classification algorithm.Through experiments,the feasibility of each module of natural language sentiment analysis based on eye tracker designed in this paper is verified.Through the above research,this paper completely simulated the research process of natural language sentiment analysis based on eye tracker,and verified the classification effect of the model in practice.The results show that the model proposed in this paper can capture the hidden information of human eyes and output the emotional tendency of human heart during the process of human eyes browsing the text.