Emotions are the backlash that a person confronts in return for an act or bearing from their surroundings. Emotion detection is a cognitive theory that advises a person’s speculations that supervise their emotions with respect to behavioral and psychological responses. The expressions posted by a person over social media are the supreme source of emotion detection nowadays. The proposed work aims at the formation of a lexicon dictionary, ESentiEmo, with 22 fine-grained emotions and 3 sentiments, consisting of 17,812 words. The proposed fine-grained dictionary has been annotated with these fine-grained emotions and sentiments. The ensemble-based deep learning models are created to effectuate the accuracy of the proposed lexicon dictionary. The proposed ensemble architecture comprises of three deep learning base architectures, namely, Long Short-Term Memory (LSTM), Support Vector Machine (SVM), and Convolution Neural Network (CNN). Our suggested framework has achieved maximal accuracy with 81.02% employing an ensemble approach coupled with maximum voting.