Research And Implementation Of The Emotional Catharsis System Based On Text Sentiment Classification

With the development of the society, a lot of people are now facing a lot of psychologicalpressure, most of them then will have some psychological emotional problems, such asirritability, anxiety, depression, anger, etc.When they have such emotional problems for a longtime, it may lead to serious psychological problems.Therefore, we need to take some effectivemethods to vent the unhealthy emotion.In real life, name-calling don’t seem civilized, butpsychological research had shown that swearing is a kind of very effective methods ofemotional outlet.As long as limited to the scope of venting and without affected others, it isappropriate to provide an environment for people to swear. The emotional catharsis systemcan automatically understand emotional catharsis text and judge the text sentiment tendency.Through the emotional dialogue, the system regulate the unhealthy emotion, so as to improveuser’s mental health.under the overall design and implementation of the emotional catharsis system, the papermainly focus on Chinese text sentiment classification. Firstly, according to the particularity ofill-natured dialogue, we collecte a lot of positive and unhealthy text from the Internet. Andconstruct an sentiment dictionary for the dialogue with the emotional values by using theexisting sentiment dictionary and the calculation of How Net semantic similarity. It cancalculate the text sentiment value by using the dictionary. Then the paper focus on Chinesetext sentiment classification under the supervised learning approach. constructing the learningsample datasets by Artificially annotating on the collected datasets of positive and negativeemotion texts. We do some compared experiment with the RBF kernel function and Linearkernel function of support vector machine, Multinomial Na?ve Bayes, Bernoulli Na?ve Bayes,Logistic regression and K nearest neighbor under the different feature selection methods andthe different number of word features. The results show that it will get the best classificationresults by using information gain as the feature selection method and support vector machineof the RBF kernel as the learning method. We then construct emotional chinese textclassification model for the emotional catharsis system to identify the sentiment of input texts.Finally, we introduce the design and construction of emotional dialogue databases.With the test and the actual use of the emotional catharsis system, the system can identifythe user’s text sentiment and give the user a corresponding reply of emotional text.It allowsthe user to vent their emotions appropriately in the affective dialogue, and achieves the resultthat we need.