Daily life patients Sentiment Analysis model based on well-encoded embedding vocabulary for related-medication text

Abstract

Millions of health-related messages and fresh communications can reveal important public health issues. New Drugs, Diseases, Adverse Drug Reactions (ADRs) keep appearing on social media in new Unicode versions. In particular, generative Model for both Sentiment analysis (SA) and Naturel Language Understanding (NLU) requires medical human labeled data or making use of resources for weak supervision that operates with the ignorance and the inability to define related-medication targets, and results in inaccurate sentiment prediction performance. The frequent use of informal medical language, nonstandard format and abbreviation forms, as well as typos in social media messages has to be taken into account. We probe the transition-based approach between patients language used in social media messages and formal medical language used in the descriptions of medical concepts in a standard ontology[21] to be formal input of our neural network model. At this end, we propose daily life patients Sentiment Analysis model based on hybrid embedding vocabulary for related-medication text under distributed dependency, and concepts translation methodology by incorporating medical knowledge from social media and real life medical science systems. The proposed neural network layers is shared between medical concept Normalization model and sentiment prediction model in order to understand and leverage related-sentiment information behind conceptualized features in Multiple context. The experiments were performed on various real world scenarios where limited resources in this case.