Health Monitoring Research For Social Media

The use of social media platform is a major feature of most people's daily life.People like to share their feelings,anecdotes,major and minor events on the Internet,including their physical condition,illness,medication and so on.At the same time,it also brings a continuous stream of data resources for natural language processing.Health has become one of the most concerned issues in the world.It is closely related to each of us and plays an increasingly important role in our life.The application of machine learning and deep learning in biomedical field is a new development trend,and will bring greater development.Social media data is generated by users themselves and has a close relationship with users.Through the mining and analysis of social media data,we can grasp some health information of users in daily life,and the happiness,anger,sadness and even health status of users can be reflected.As long as we capture this information and make good use of social media data,we will be able to "monitor" our health status and timely feed back the changes of our mentality or illness.In this paper,social media data as the research object,through natural language processing method to mine information related to user health,so as to realize the monitoring of user health.(1)Aiming at the standardization task of social media,this paper proposes the Bio DRA(Bio BERT based model with Dilated RNN and Attention mechanism)model to transform the informal description of adverse drug reactions in social media into professional terms.The model adopts the pre training model of vertical domain and character level embedded supplementary text representation,and obtains the context information through the bidirectional extended recurrent neural network.Finally,it improves 2-3 percentage points compared with the best score on the public data set.(2)After preprocessing the data,this paper selects two health-related concepts: happiness and depression.Happiness is an abstract concept.We introduce two fine-grained concepts "agency" and "social" related to happiness."Agency" refers to the sense of happiness under the action we can control,and the "social" label means that the moment of happiness involves other people.In this paper,a Senti-BSAS(BERT embedded bi-lstm with Self-Attention mechanism along with the Sentiment computing)model is proposed for the classification task of the two tags.On the basis of BERT(Bidirectional Encoder Representation from Transformers)and LSTM(Long Short-Term Memory),the tag information is integrated into the affective dictionary,In order to improve the performance of the model,the tag related external emotional resources are used to express emotions.(3)Because of its characteristics,depression is often ignored in the early stage of the disease,but social media has opened a window of communication for patients with depression.This paper proposes a Bio-CSD(Bio BERT based model with Char embedding and Sentiment computing for Depression research)model for depression "monitoring" classification task in social media.Combined with the characteristics of the first two models,it uses the vertical domain pre training model,character level embedding and self-attention mechanism,and introduces new emotional resources to enrich emotional expression.The experimental results show that the model achieves the best results,which provides the possibility for the discovery of depression through social media data.