A comparative study on word embedding techniques for suicide prediction on COVID-19 tweets using deep learning models

COVID-19 caused a pathetic situation worldwide which led to public health crises, economic crises, employment losses, and mental anxiety. Social media websites are being inundated with reports on the virus, which has led to a variety of perspectives, thoughts, and emotions being expressed and experienced by social media sources. Taking advantage of the amount of information available, an analysis of sentiments user opinions expressed can be done on social networks. Sentiment Analysis is widely used in social media platforms for understanding the user’s expressions and sentiments. In this work, we extricate the information from Twitter utilizing search words, a python API called tweepy, pre-process it, and perform the word embedding process. The Word Embedding process is the replacement for the one-hot encoding technique, which converts the given into the form of vectors by tokenizing them as words and also spotting the relation among the words. Word Embedding Techniques such as Word2Vec, Glove, and FastText are used to convert the text into vectors which are then fed to the Artificial Neural Network (ANN) Models for training. Valance Aware Dictionary and sentiment Reasoner (VADER) are used to detecting suicide propensity in tweets as positive, negative, and neutral after which the user can be notified and solutions are provided. As a result, we will be able to figure out the user’s suicidal feelings and emotions during the pandemic situation. This research is used for comparing different word embedding techniques and predicting the suicide inclination of the tweets using the word embedding vectors and the neural network models.