Emotion Detection Using Machine Learning Algorithms: A Multiclass Sentiment Analysis Approach

Abstract

Due to the availability of internet technology and social media platforms, people are well connected with the world. People express their feelings on social media platforms such as Facebook, twitter, reddit, and Instagram. Feelings can be expressed with text, images, or videos. Due to widespread usage of social media platforms on the internet, a large amount of unstructured data is being generated. To understand the human psychology and emotion state, this data needs to be processed to identify text polarity with the sentiment analysis and emotion detection, respectively. This paper gives a brief introduction about sentiment analysis, emotion detection, and machine learning algorithms performing well for this task. We used naïve bayes, logistic regression, XGBoost, and linear support vector classifiers for the model building. Out of these classifiers linear support vector classifier gave high accuracy of 72.80% and performed well to predict the emotion of Marathi or English-speaking user at runtime.