A Conditional Sentiment Analysis Model for the Embedding Patient Self-report Experiences on Social Media

Abstract

Getting accurate, honest, reliable and credible minute insight is the most crucial objective of conducting medical and pharmaceutical research on social media. Nowadays, healthcare manufacturing companies use Sentiment Analysis (SA) to identifying the misleading of patients self-report experiences and shared medical information on social media. As a target level of analysis, a set of medical components in each document (post, message, tweet, etc.) have a semantic formalism which, similar to a dependency parse in the whole space of analysis regarding the time axes. However, Time property is been substantially very important allowing more real-time personalization to efficiently detect patient emotional state and what may be suffering from. Specially, when an irregular sentiment towards drugs or set of events may cover. In this paper, we aim at defining a conditional Sentiment Analysis model which summarizes sentiment information looking at the historical data towards dependent entities for yielding short or long-term predictions based on quantifying exactly what change is. This model hybrid an unsupervised biomedical concept extraction with autoregressive time series modelling. This hybridization aims at online updating the model by smoothing and extracting new relevant target features when deals specifically with newly emerged diseases, medical events, Drug issues and potential side effects. The evaluation results on a real pharmaceutical industry and healthcare tweets show that our proposed oriented-context method performs better than existing models.