Sentiment Tendency Recognition Method Of Depression Text Based On Sentiment Word Extraction

Depression is a common mental illness in today’s life.In recent years,with the increasing pressure of life,depression is also increasing year by year,and the suicide rate is also on the rise.Such a serious mental illness is harming people’s physical and mental health,and even has a negative impact on the society.Therefore,it is urgent to find out whether there is a tendency to depression and treat it as soon as possible.With the rapid development of social networks,more and more depression users to express themselves in the social networking platforms such as weibo,by statements to record their own psychological change,the illness condition and the next step may take extreme behavior.Therefore,it is of great significance to identity depression tendency based on online comments of depressed patients.To solve the above problem,this paper proposes a method to identity the sentiment tendency of depression texts based on sentiment word extraction.The specific research content is as follows:(1)Constructing depression lexicon based on depression comments and word vectors,which provides an important basis for identifying depression tendency.Firstly,TF-IDF algorithm is used to obtain the frequency of words,and labelling their sentiment intensity to obtain high intensity and high frequency sentiment seed words;Moreover,artificial screening is carried out to obtain high frequency behavior seed words.Then,the candidate words is extracted by word vectors extraction from the comments of patients with depression through BERT pre-training model.Semantic similarity between the seed words and the candidate words was calculated to obtain the sentiment words and behavior words of depression repectively.Finally,the correlation degree between sentiment words and behavior words is obtained by means of statistics and calulation of point mutual information,so as to obtain the corresponding relationship between behavior words and sentiment words.(2)Constructing the depression tendency recognition model based on the depression lexion and Transformer model,which achieves the purpose of depression tendency recognition.Firstly,the sentiment feature information of depression dictionary is learned by self-attention mechanism,and the semantic feature information of sentences is extracted by Transformer model.Then,the sentiment feature information of dictionary and the semantic feature information of the sentence are fused to obtain the fused feature matrix,and the feature representation is obtained by calculating the feature matrix with pooling layer.Finally,feature representation is calculated by linear layer and Softmax layer to obtain the classification result of depression comment text sentiment orientation.The experimental results show that the method based on sentiment words extraction can not only effectively mine the sentiment words and behavior words of depression,but also improve the accuracy of sentiment orientation recognition of depression comment text.Figure[17] Table[15] Reference[80]...