Sentiment Analysis Model using Text and Emoticons for Pharmaceutical & Healthcare Industries

Abstract

Background: Nowadays, pharmaceutical and healthcare industries apply sentiment analysis (SA) to analyze customers' feedback on pharmaceutical products. SA uses natural language processing (NLP) to recognize, assess and analyze the pharmaceutical data to extract valuable insights about the sentiment it conveys. Over the past two decades, in addition to text, emoticons have become a conventional way of expressing one's sentiments. The researchers utilizing textual data for SA have largely ignored emoticons due to their complexity and insufficient resources. Aim of the study: Necessary to include emoticons in SA to capture the emotion expressed by pharmaceutical data. However, this research develops an algorithm that performs SA on text and emoticons. Methods: To highlight the significance of emoticons, the analysis has been conducted on text-only and text-and-emoticon data, using deep learning (DL) and machine learning (ML) methods on multiple datasets. The features – term frequency-inverse document frequency (TF–IDF), emoticon lexicons, and a bag of words (BoW) are taken for simulation. Results: Simulation of text features using BiLSTM provides better accuracy, between 85% and 90%, compared to other methods. The precision lies between 70% and 90%. The recall is lowered to 64%. The sentiment the emoticon conveys outweighs that of the text associated with it. In addition, the DL algorithms outperform the ML methods. The overall results indicate that considering emoticons along with text positively impacts SA.Further, it is noted that the DL model marginally outperforms all of the ML algorithms used. Conclusions: This research analyzed the augmentations to the implementation, including multiple enhancements during the pre-processing stage, the inclusion of emojis, and the usage of a bi-long short-term memory network (BiLSTM) in the DL model. The experimental results conclude that better measures are obtained using the DL method.