Deep Neural Network Approach for Sentiment Analysis of Tweets Related to Sports Concussion

Abstract

Concussion and Traumatic brain injuries (TBI) are serious injuries that impair the normal functionality of a person’s brain. Symptoms can include confusion, disorientation, loss of consciousness, memory lost, and in more sever situations fatality. It is reported that 39% of children (ages 10-18 years old) who visit the hospital due to a sports-related head injury were diagnosed with concussion and 24% with the possible concussion [1]. In order to bring awareness about the seriousness of the TBI to the attention of the policy makers, a neural network based sentiment analysis ensemble system that automates the process of gathering the opinion of the general public is designed. A preprocessing pipeline is proposed that embeds various word-level features into a single concatenated vector. Input vectors are processed by varying Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) Networks. The proposed ensemble system achieves an evaluation score of 62.71% based on its precision and recall, and compares well with other state-of-the art systems.