Medical Sentiment Analysis using Social Media: Towards building a Patient Assisted System

Medical Sentiment Analysis using Social Media: Towards building a Patient Assisted System Shweta Yadav, Asif Ekbal, Sriparna Saha, Pushpak Bhattacharyya Department of Computer Science and Engineering Indian Institute of Technology Patna Patna, Bihar, India {shweta.pcs14, asif, sriparna,pb}@iitp.ac.in Abstract With the enormous growth of Internet, more users have engaged in health communities such as medical forums to gather health-related information, to share experiences about drugs, treatments, diagnosis or to interact with other users with similar condition in communities. Monitoring social media platforms has recently fascinated medical natural language processing researchers to detect various medical abnormalities such as adverse drug reaction. In this paper, we present a benchmark setup for analyzing the sentiment with respect to users’ medical condition considering the information, available in social media in particular. To this end, we have crawled the medical forum website ‘patient.info’ with opinions about medical condition self narrated by the users. We constrained ourselves to some of the popular domains such as depression, anxiety, asthma, and allergy. The focus is given on the identification of multiple forms of medical sentiments which can be inferred from users’ medical condition, treatment, and medication. Thereafter, a deep Convolutional Neural Network (CNN) based medical sentiment analysis system is developed for the purpose of evaluation. The resources are made available to the community through LRE map for further research. Keywords: Medical sentiment, Medical blog, Machine learning, Deep learning, Convolutional Neural Network 1. Introduction Attention towards sentiment analysis has been flourishing over the last two decades because of the immense popu- larity of social media. The phenomenal rise in blogging trend is observed in health communities such as medical forums which are swamped by millions of users (many of whom are patients) seeking for health-related information, sharing medical problems or experiences and opting for in- formational support or opinions from the other users (pa- tients, health-professional or doctors). These self-narrated texts provide a platform to peek into a blogger’s state-of- the-mind for several reasons: (i) the subjective nature of the contents generated by blog- ger’s; (ii) the temporal aspect of the blog which can be formed into thread; (iii) the abundance of blog data which allows cumulation of opinions, sentiments and thoughts in a very wide spectrum. Medical sentiment analysis has its major applications in as- sessing the clinical records and in providing an automated decision support system for health professional. Accord- ing to the study conducted by the Pew Internet & Ameri- can Life Project1, almost 80 percent of Internet users in US have explored health-related topic online. More often, peo- ple look for the information about specific medical problem (63%) over the internet. Nearly 47% of the users search for the medical treatment or procedure in the internet. With such a tremendous amount of freely available medical texts in the web, it is necessary to harness the crucial and im- portant information. Analyzing these texts by capturing the sentiments is helpful because opinions are central to almost all human activities and are key influencer of our behaviors. Although, …

Content truncated to 500 words