Research Of Depression Recognition Based On Micro-blog Text And Deep Learning

Depression is a typical chronic mental illness.Its distinctive features are emotional negative tendencies and behavioral suicidal tendencies.Reports published in recent years show that 4.3% of the world's population is currently suffering from depression.Among them,more than 50 million people suffer from depression in China,which accounts for 4.2% of the total population.The prevalence rate reached6.1%,and the suicide rate accounted for 50% of the total suicide population.However,due to personal and social influence,the current recognition rate of depression is less than 10%.How to improve the recognition rate is an urgent problem to be solved in the treatment of depression in China.With the development of social network platform,more and more depressive patients take micro-blog as a way to express themselves.These daily life recorded by micro-blog contain a large amount of patient information.It provides a new method for the identification of depression.On this basis,a depression recognition algorithm based on micro-blog text and deep learning is proposed in this paper,which not only effectively avoids the problems existing in the current depression recognition,but also provides support for medical staff to actively discover and rescue patients.The main contents of this paper include:(1)Construction of a dictionary database for depression: Analyzing the common characteristics of emotion and behavior in depression micro-blog.By using two semantic similarity algorithms and combining general knowledge base with experimental corpus,a dictionary database of depression is constructed which covers emotional dictionary,emoticons dictionary,keyword dictionary and so on.It makes up for the vacancy of the dictionary in this field.(2)Pretreatment of data:The experimental corpus is formed by de-privacy,word segmentation and de-noising of original data.The dictionary features,semantic features and extended features closely related to depression are extracted,which provides support for subsequent recognition.(3)The construction of the depression recognition model:Depression recognition is transformed into micro-blog text classification.Shallow support vector machine and deep convolution neural network algorithms are used for experiments respectively.According to the characteristics of depression micro-blog,the algorithm is further improved.And a dual-input convolutional neural network algorithm compatible withmulti-features is proposed.Experiments show that the recognition rate is effectively improved.(4)The conversion of patients' micro-blog to electronic medical records: The micro-blog identified as depressive patients is transformed into electronic medical record form by the tool,which facilitates the research and analysis of medical staff and provides support for follow-up rescue.To sum up,this method of depression recognition makes full use of natural language technology and social media,which breaks the traditional patient's active treatment mode.It has important significance for rapid identification and early warning of depression.