The recent trends in Artificial Intelligence (AI) are all pointing towards the singularity, i.e., the day the true AI is born, which can pass the Turing Test. However, to achieve singularity, AI needs to understand what makes a human. Emotions define the human consciousness. To properly understand what it means to be human, AI needs to understand emotions. A daunting task, given that emotions may be very different, for different people. All these get even more complex when we see that culture plays a great role in expressions present in a language. This paper is an attempt to classify text into compound emotional categories. The proposal of this paper is identification of compound emotions in a sentence. It takes three different models, using Deep Learning networks, and the more traditional Naïve Bayes model, while keeping the mid-field level using RAKEL. Using supervised analysis, it attempts to give an emotional vector for the given set of sentences. The results are compared, showing the effectiveness of Deep Learning networks over traditional machine learning models in complex cases.