Depression Recognition Of Weibo Users With Text Categorization

Depression,as one of the world’s three major health “killers”,is mainly manifested in significant and persistent low mood,pessimism,world-weary,frailty,etc.Severe cases can lead to suicidal behavior.In the face of the increasingly serious depression situation in China,attention should be paid to active prevention and treatment.The current traditional methods to identify depression tendencies are mainly manual testing,including depression self-assessment scale,psychological counseling,follow-up surveys,etc.These methods are generally slow and have some lag.On the other hand,although many applied studies on text classification have been realized in the previous years,there are very few studies on the use of text classification in China to detect depression tendencies.Based on this background,this paper uses text classification to construct machine learning models and identify the depression tendency of Weibo user comments,which can effectively improve the efficiency of depression detection to detect depressed patients in time,provide new ideas to timely intervention treatment and the prevention of the occurrence of tragedy.It has a certain positive significance.The main work and innovations of this paper are as follows:First,this paper uses targeted reptiles to obtain the original data of Weibo comments and manually label the data type.Through the reptile program,we obtain comment data from "Yiyuzheng Chaohua”,"Zoufanzou Chaohua" and the last comment of the blogger"Zoufan" in her lifetime,and whether depression is artificially labeled.Second,combined with statistical and psychological theory,the dictionary of depression is constructed.Based on the statistical results of the word frequency of depressed users’ comment texts,combined with the research of psychology in depression,this paper divides the comment of depressed users into five aspects: mental state and emotional expression,interpersonal relationship and social support,sleep quality,disease and treatment,and thoughts of suicide.Based on the five major aspects,a new depression dictionary is generated to provide data support for the study of depression tendency recognition.Third,this paper is based on machine learning for the identification of depression tendencies.In this paper,the traditional artificial recognition of depression is innovated and the identification of depression tendency is converted into a two-classification problem.According to TF-IDF and Word2 vec methods to extract feature vectors,we use decision tree,support vector machine,naive Bayesian classifier,and long-short-term memory(LSTM),which has become popular in the field of deep learning in recent years,to construct a text classification model of depression tendency.The experimental results show that the LSTM model has significantly better classification performance than the other three traditional machine learning models,with an accuracy rate of 97%.Therefore,it is feasible to apply this model to depression recognition research of Weibo comments.